Internet Published Manuscripts 0:1-51 (1999)

Original Article In the Theater of Dreams: Global Workspace Theory, Dreaming, and Consciousness

Donald J. DeGracia Department of Physiology, and the Center for Molecular Medicine and Genetics, Wayne State University, Detroit, Michigan

Utilizing the Global Workspace System of Baars

(1988), this paper compares conscious and uncon-scious processes across waking, nonlucid, and lucid dreams. Sleep psychology can display gross functional dissociation between perceptual and cognitive con-sciousness. We utilize this observation to develop models of sleep experience and dream generation. These models accommodate Hunt’s (1989) “multiplicity of dreams”, as well as the intrinsic variation of percep-tual and cognitive activity during dreaming. Lucid dreams are suggested to result from the presence of a skill-based mental set, the lucid dream context, which allows voluntary interaction with the spontaneous dream process. Our view of dreaming provides an explanation of the tendency of lucid dreams to either fade or revert to nonlucid dreams. Neurobiological considerations lead us to hypothesize that, in the sleeping brain, a re-versal of information flow from medial temporal lobe mnemonic structures to thalamocortical perceptual cir-cuits imparts parameterization to dream perceptual con-sciousness. A consequence of our thinking is that dreaming results in a “mental recombination” of cerebral information networks, which contributes to the ability of waking consciousness to generate novel and adaptive responses. © 1999 Donald J. DeGracia

Key Words: consciousness, unconscious, contexts, dreams, Global Workspace, lucid dreams, medial tem-poral lobe memory, thalamocortical circuits, recombina-tion

INTRODUCTION

Although we have seen stunning progress in the cognitive neurosciences in the past decades, our under-standing of dreaming has not shared proportionally in these advances. Substantive omissions in dream theory have made the realization of a robust science of dream-ing elusive: (1) there is no consensus for comparing

waking and dreaming psychology; dream theoreticians are divided over the issue of whether dream psychology is continuous or discontinuous with waking psychology (Antrobus, 1986; Purcell, Moffitt, & Hoffmann, 1993), (2) theoretical approaches to dreaming lack unity; there is still no clear conception linking biological, interpre-tive, and content analysis approaches to dreaming (Kahn, Pace-Schott, & Hobson, 1997; Moffitt, Kramer & Hoffmann, 1993), and (3) the study of lucid dreams has proceeded in relative isolation from the study of nonlucid dreams, and lucid dreaming data has had minimal impact on either dream theorizing or models of waking (Kahan & LaBerge, 1994; Purcell et al., 1993). The purpose of this paper is to explore the implications of a comparison of the phenomenology of gross subjec-tive experience across waking and dreaming. The ap-proach we utilize here begins to address the above theo-retical issues.

Two assumptions seem to have played a central role in the promulgation of the above theoretical shortcom-ings. The first assumption has been recently formulated by Llinás and Paré (1991, 1996) in terms of thinking of the function of the central nervous systems (CNS) as an open or a closed system. To quote:

Contract grant sponsor: None. Done for Free.

*Correspondence to: Dr. D.J. DeGracia, Department of Physiology, Wayne State University, 4116 Scott Hall, 540 East Canfield Ave., Detroit, MI 48201. E-mail: ddegraci@med.wayne.edu Published online at http://www.geocities.com/ddegraci/index.html

2 DeGracia

“…we regard as fundamental the issue of whether the functional organization of the CNS is to be considered to have a closed or open architecture. An open system is one that accepts inputs from the environment, processes them, and returns them to the external world as a reflex regard-less of their complexity…This view, which is still perva-sive, explains nothing about the function of the CNS or the invariance of its function among individuals….If we opt for the closed-system intrinsic hypothesis, it follows that the nervous system is primarily self-activating and capable of generating a cognitive representation of the ex-ternal environment even in the absence of sensory input, for example in dreams...From this perspective, cognition as a functional state may be considered as an a priori property of the brain…” (Llinás & Paré, 1996, pg. 4-5)

The assumption of an open system view of the brain

may be the basis of discontinuity views of dreaming, e.g. that dreams are fundamentally different from wak-ing experiences (Hartmann, 1973; Hobson & McCarley, 1977; Rechtschaffen, 1978). An example of how an open system view can bias dream research is the idea that seeing in dreams is akin to waking visual imagery but not visual perception (Kerr, 1993); retinal informa-tion clearly has nothing to do with what is seen in dreams, leading to the conclusion that seeing in dreams cannot be true visual perception. However, if we adopt a closed view of the CNS, then visual perception and imagery are two innate properties of the brain, and there is no reason in principle why both cannot simultane-ously occur in the sleeping brain just as they do in the waking brain. Although a closed system view implies a fundamental continuity between waking and dreaming, because the sleeping brain undergoes gross functional changes, we expect such changes to be reflected psycho-logically. The view we develop here is a composite continuity/discontinuity view.

A second major assumption, captured in a quote from Purcell, et al., (1993), is that there has been “…a widespread view that waking and sleeping correspond to conscious and unconscious processes.” That is, to the external observer, a sleeping person appears uncon-scious, hence it has been natural to associate forms of sleep mentation such as dreaming with unconscious mental processing. Likewise, a waking person is clearly conscious, and so waking is associated with conscious mental processing. The distinction of con-scious/unconscious is particularly strong in clinical in-terpretive traditions (Freud, 1900), and generally ig-nored in biological models of dreaming, or not treated in

a systematic fashion (e.g. Kahn et al, 1997). However, during lucid dreaming for example, the dreamer is con-scious, not only of the dream environment, but of the fact that they are experiencing a dream. If dreaming is associated with unconscious processing, how is it that lucid dreamers can display forms of consciousness equivalent to those expressed during waking (Kahan, LaBerge, Levitan, & Zimbardo, 1997; LaBerge, 1985; Purcell, et al., 1993)? Similar considerations apply to nonlucid dreaming where, for example, some research-ers point to the recall of dreams as evidence that dreams involve conscious processing (Fiss, 1983). A related issue we shall address below is the claim that dreaming psychology is deficient relative to waking psychology, which is again difficult to sustain as a formal feature of dreams in the face of the phenomenology of lucid dreaming.

In this paper, we adopt the closed system view of CNS function advocated by Llinás and Paré, and explic-itly assume that the repertoire of innate functions ex-pressed by the waking brain is also expressed in the sleeping brain. However, we take this view a step fur-ther because the phenomenology of sleep mental experi-ence indicates that the human CNS generates two global conscious constructs: (1) a self, and (2) representations of an external perceptual environment. With respect to dreaming, these representations correspond to the dreamer - the person undergoing the dream experience, and to the dream environment - the perceptual setting in which the dream occurs. As we shall elaborate, exam-ples of functional dissociations between the dreamer and the dream environment exist, and these two representa-tional constructs appear to vary independent of one an-other.

The main tool we will utilize for comparing the self and the perceptual environment across waking and dreaming is Bernard Baars’ Global Workspace (GW) model (Baars, 1988). Our analysis will not only be an application of the GW model to dreaming, but an exten-sion of this model to accommodate the types of mental experiences that occur during sleep. The GW definitions of “conscious” and “unconscious” can be used inde-pendent of the global state of the brain. Therefore, this model provides us with a consistent and systematic framework for discussing conscious and unconscious processes involved in representations of the self and the perceptual environment across waking and sleep. The GW model, via the notion of “context,” also provides a means for conceptualizing higher order forms of mne-

Global Workspace and Dreaming 3

monic organization in the nervous system, which is par-ticularly important for understanding the role of memory in dreaming. Clarifying the relationship of memory and dreaming serves two purposes: (1) it will begin to bridge the biological, interpretive, and content analyses ap-proaches to dreaming, and (2) it will allow us to present a neurobiological model of how the dream perceptual environment is formed from mnemonic structures in the sleeping brain. Our GW analysis will also allow us to propose a new hypothesis of the function of dreaming: dreaming is a form of “mental recombination” that indi-rectly serves to enhance the flexibility and adaptability of the waking brain.

FUNCTIONAL DISSOCIATION OF THE SELF AND PERCEPTUAL ENVIRONMENT IN CONSCIOUSNESS DURING SLEEP

It is now clearly recognized that mental experiences occur throughout the entire sleep-wake cycle (Bosinelli, 1995; Foulkes, 1967; Hobson, 1988). We begin by con-sidering the conscious aspects of sleep experiences. We will define below our usage of the terms “conscious” and “unconscious”. For the moment, the term “con-scious” can be taken in its common sense usage as, for example, you, the reader, are conscious of these words, or your environment, or of yourself and your thoughts, etc.. Three major forms of sleep experience contain definite conscious content (Foulkes, 1962): dreams, hypnagogia, and thought-like mentation. The conscious aspects of dreams, hypnagogia and thought-like experi-ences can be conceptualized in dual terms in which there is a dreamer and a perceptual environment.

In dreams, there is a subject, the dreamer, a con-scious person to whom the dream happens. The dream occurs within a vivid perceptual environment in which all perceptual modalities can and do occur. The con-scious aspects of the dreamer can be divided into: (1) the conative aspects of conscious cognition such as the dreamer’s thoughts, attention, volitions, and imagings (Purcell et al., 1993), and (2) the dreamer’s conscious perception of the dream environment. The dreamer and the dream environment can be separated on the basis of intention: the conative aspects of the dream are willed by the dreamer, the perceptual aspects are not. That is, the dream environment is not in any obvious way inten-tionally created by the dreamer. We take this to repre-sent a fundamental dissociation between cognition and perception in dream consciousness. A major purpose of this paper is to address such a dissociation. The concept

that dreams contain a “plot” or “narrative” has also guided some investigators’ conceptions of dreaming (Cipolli & Poll, 1992; Foulkes, 1978; Kuiken, Neilson, Thomas, & McTaggart, 1983; Kuper, 1979). The dreamer’s interaction with the dream environment can take two forms: the dreamer can be immersed within the dream perceptual environment (dreamer-as-actor dreams), or may appear to be outside the dream envi-ronment, observing it as if watching television (dreamer-as-observer dreams) (Foulkes & Kerr, 1994; LaBerge, 1985).

In this paper we are not interested in focusing on the controversy surrounding NREM versus REM dreams. Some of this controversy stems from conflicting defini-tions of these experiences (Kahn et al., 1997; Okuma, 1992). We note here that evidence generally indicates that NREM dreams are simpler than REM dreams along both cognitive (Cavallero, Cicogna, Natale, Occhionero, & Zito, 1992) and perceptual dimensions (Antrobus, Hartwig, Rosa, Reinsel, & Fein, 1987; Antrobus, Kondo, Reinsel, & Fein, 1995) and occur with a low frequency compared to REM dreams (Dement & Kleit-man, 1957; Foulkes, 1962). In the scope of our think-ing, NREM dreams result from the same mechanisms as REM dreams (Cicogna, Cavallero, & Bosinelli, 1991) although lower levels of cerebral activation may be re-lated to their decreased complexity (Foulkes, 1967). When we refer to dreams throughout, it is implied that our logic applies to both REM and NREM dreams.

Hypnagogia, predominant at sleep onset (or during the sleep-to-wake transition, where it is called hyp-nopompic imagery), is associated with visual percep-tions, although auditory and somatic perceptions can also occur (Schacter, 1976; Schneck, 1968). However, hypnagogic perceptions are less structured than dream perceptions, sometimes surreal in character (Gillespie, 1989), lacking “narrative” elements, and the visual im-agery may appear as simple static “snapshots” (Mavro-matis, 1987). Hypnagogic perceptions are observed by a relatively passive dreamer, and occur as if external to the dreamer, similar to dreamer-as-observer dreams (Mavromatis, 1987).

Thought-like mentational experiences, termed “NREM mentation” because of their association with NREM awakenings (Foulkes, 1967; Hobson, 1988), more resemble waking thinking and lack the vivid per-ceptual modalities associated with dreaming. Further, the form of this thinking is banal and repetitive, the con-tent reflecting mundane waking concerns. To quote

4 DeGracia

Hobson (1988, pg 143), NREM mentation is “nonpro-gressive…the mind seems to be running in place.”

An identical dreamer/environment dichotomy holds during lucid dreams. This dichotomy is particularly evi-dent at the onset of wake-initiated lucid dreams (WILDs), and during the process of fading from a lucid dream (both phenomena are reviewed in LaBerge and DeGracia, 1999). The following example from D.J.D.’s personal dream journal illustrates a functional dissocia-tion between himself as the dreamer and the perceptual environment, during the onset of a WILD:

“Was aware of being in my body. I wanted to go out.

Laid concentrating, desiring to leave…soon, I wasn’t aware of anything around me. In my imagination, I imag-ined flying off, but got pulled back again. This happened twice. Then, I dove off my bed straight downwards. I was moving straight downwards in the void. Far below me in the darkness I saw a square floating. In the square I could see colors, like a scene was inside the square. I stretched to grab this square and my arms stretched far below me like Plastic Man, and I grabbed the floating square. I pulled it up over me like putting on a pair of pants, and was thinking to myself, ‘I’m not gonna let this one go!’ I stepped into this square and was all of a sudden some-where! I was very surprised! I was in what looked like a high school hallway standing in line with people going into a room…”

This quote describes the contents of D.J.D.’s con-

sciousness during the transition from laying awake in bed to appearing in a dream fully lucid. D.J.D. lost awareness of the external perceptual environment, leav-ing him momentarily in a state devoid of any external environment. However, his conscious cognition, includ-ing internal speech, self-reflective and imagery capabili-ties, and the ability to form episodic memories remained intact. He then imagined the sensations of flying up-wards. This imagery transitioned almost imperceptibly to perceptual sensations occurring external to him, form-ing a “minimal” perceptual environment (LaBerge & DeGracia, 1999; see also section 7). The minimal envi-ronment, termed “the void,” was perceived as a dark space (e.g. visual-spatial depth perception devoid of vis-ual content), through which he could move (kinesthesis), although there was no sense of a body image (somesthe-ses) or any other perceptual modalities. Next, within this “void,” a form of hypnagogic intrusion (Hunt, 1989) occurred and D.J.D. saw, in the apparent distance, a square containing colored imagery floating. By volun-

tarily willing his arms to reach out and grab at the square, this somehow caused a bizarre visual appearance of arms, and the kinesthetic sensation of reaching at the square and pulling it upwards. Upon “putting on” the scene inside the square, D.J.D. then abruptly found himself in a fully-formed dream perceptual environment. This environment, resembling a high school hallway, contained the full repertoire of perceptual modalities, and was very similar in quality to waking perception. This environment was unfamiliar to D.J.D., had nothing to do with his conscious thoughts and intentions, and appeared spontaneously. What is striking is that the dreamer remained intact cognitively while the percep-tual environment underwent such drastic variation.

It is precisely such experiences that have inspired the approach we develop in this paper. To our knowl-edge, no attempt has been made to quantitate the fre-quency of conscious WILD transitions amongst experi-enced lucid dreamers. D.J.D. has remained conscious across the wake-sleep border in 22 out of 103 (21%) lucid dreams, or 22 out of 45 (49%) WILDs. Because of differences in styles of lucid dreaming, we expect such statistics would vary among experienced lucid dreamers (LaBerge & DeGracia, 1999). It is also interesting to note the onset of perceptual modalities in this WILD transition: visual-spatial depth and kinesthesis during the “void” phase, followed by rich visual, somatic and auditory content during the “dream” phase. The modali-ties associated with the “void” phase correspond to con-ceptions of the dorsal visual pathway linking occipital and parietal lobes, believed to encode visual-spatial depth and motion (Bertenthal, 1996; Ungerleider & Haxby, 1994). The subsequent “dream” phase entailed simultaneous onset of modalities associated with the ventral visual pathway in inferior temporal lobe which encodes visual form and color perception (Ungerleider & Haxby, 1994), along with auditory and somatosensory modalities. This pattern of dream onset is essentially the opposite of that associated with lucid dream fading (La-Berge & DeGracia, 1999; LaBerge, DeGracia, & Zim-bardo, 1999). Although evidence of differential devel-opment of sleep in cortical areas exists (Pigarev, Noth-durft, & Kastner, 1997) our phenomenological observa-tions suggest a very specific pattern of cortical activa-tion and deactivation associated with the formation and loss of the dream perceptual environment.

This brief review of sleep experience phenomenol-ogy indicates several important generalizations:

Global Workspace and Dreaming 5

self

PE

Dreamer as actor

self

Thought-like mentation

PE

Dreamer as observer

self

PE

Hypnagogia

self

self

PE

self self

PE

self

PE

Awareness ofwaking perceptual

environment

No awareness ofany perceptualenvironment

Formation of“minimal” perceptual

environment

Awareness ofdream perceptual

environment

B

A

Fig. 1. The structures of sleep conscious experiences in terms of a dual representation of the dreamer (self) and the dream perceptual environment (PE). [A]. The normative types of sleep experience. The transition from thought-like mentation to dreamer-as-actor dreams shows a series of increasing activity in both generators in terms of the com-plexity of the representations and their interactions. [B]. Illustration of changes in consciousness during a wake-initiated lucid dream (WILD). See text for details.

The CNS generates consciousness of two global and functionally distinct representations: a self, and a per-ceptual environment.

The complexity of either of these representations can vary. A spectrum of complexity is observable in the perceptual environment, ranging from none in the case of NREM mentational experiences, to simple, static im-agery in the case of hypnagogia, to complex multimodal environments in the case of dreams. A similar spectrum of complexity of the dreamer is observed, ranging from the banal thinking of NREM mentational experiences, through the relatively passive observer of hypnagogia and dreamer-as-observer dreams, to the actively partici-pating dreamer of nonlucid dreams, and culminating in the waking-like cognition of the lucid dreamer.

The relative complexity of the dreamer and percep-tual environment appears to vary independent of one another. In the case of mentational experiences, a dreamer exists, but a perceptual environment does not, indicating that conscious cognition can occur in the ab-

sence of perceptual consciousness. Although the per-ceptual environments of nonlucid and lucid dreams are similar (Gackenbach, 1988), the conscious cognition of the dreamer is different between nonlucid and lucid dreams (a distinction we elaborate more fully below), indicating that the complexity of the dreamer can vary while the perceptual environment stays relatively con-stant. The WILD example above shows a cognitive constancy of the dreamer while the perceptual environ-ment varied drastically.

The degree of interaction between the dreamer and the dream environment can vary, ranging from no inter-action in the case of NREM mentation, through simple observation of the environment by the dreamer in hyp-nagogia and dreamer-as-observer dreams, to a complete perceptual and motoric immersion of the dreamer within the environment in fully formed nonlucid and lucid dreams.

In Figure 1, we have developed a simple graphical notation to depict the conscious representations of self

6 DeGracia

and the perceptual environment, and their relationship, in the sleep experiences discussed above. The dreamer and the perceptual environment are depicted simply as bubbles, the relative size of which indicates the com-plexity of the representations. The interactions are shown by the relative locations of the bubbles: observer-like interactions show the bubbles separated, immersion interactions show the self embedded in the perceptual environment. Depicting conscious sleep experiences as in Figure 1A suggests a formal relationship between these forms of experience. From left to right in Figure 1A we observe a progressive increase in the complexity of both representations and the interaction between them, indicating that a spectrum of activation relates forms of sleep experience, which we address more pre-cisely in section 5.2. Such a relationship parallels the global activation state of the brain during the sleep stages predominantly associated with each of these types of sleep experience: REM dreams > stage II sleep onset hypnagogia > stage III/IV thought-like mentation (Ga-staut, 1969). Figure 4B shows a graphic representation of the WILD transition described above. Interestingly, this transition seems to approximate a NREM to REM transition with respect to the increasing complexity of the perceptual environment, although the representation of the dreamer remains constant.

The above view of sleep experience has relevance to the Activation/Input/Modulation (AIM) model presented by Hobson and colleagues (Hobson & Stickgold, 1995; Kahn et al., 1997). The AIM model provides phase space representations of three variables: activation, A (representing brain activation as measured by EEG power), input/output gating, I (a measure of external to internal input as measured by sensory thresholds or mo-tor output), and neurotransmitter modulation, M (a measure of the global neurotransmitter state of the brain as assessed by the ratio of aminergic to cholinergic brainstem neurotransmission). The AIM model plots states of consciousness as a single AIM triad. It seems reasonable to ask if a single triad of AIM variables can account for the apparently independent activities of the dream environment and the dreamer? Also, input/output gating is not a unitary phenomena; sensory input is a perceptual function, effector output is a cognitive func-tion mediated by the dreaming self. If we compare cen-tral activity during REM, then input and output are vastly different. For example, primary visual cortex is virtually silent during REM (Braun, Balkin, Wesensten, Gwadry, Carson, Varga, Baldwin, Belenky, & Hersco-

vitch, 1998), motor cortex is active by several criteria (Hess, Mills, Murray, & Schriefer, 1987; Steriade, Iosif, & Apostol, 1969; Porte & Hobson, 1996), but is inhib-ited from peripheral expression (Hishikawa & Shimiz, 1995; Steriade, 1992). These changes represent differ-ent perceptual and cognitive activities, respectively, and further suggest the need to dichotomize the AIM por-trayal of sleep experiences.

In summary, the phenomenologies of the variety of sleep experiences are evidence for a functional dissocia-tion of conscious representations of the self and the per-ceptual environment within the CNS during sleep. We utilize this observation, in conjunction with the GW model, to develop a framework that will allow us to sys-tematically compare conscious and unconscious proc-esses across dreaming and waking.

The arguments presented in Llinás and Paré (1991), the essence of which is quoted above, indicate that a third gross function occurs in the awake nervous system: the parameterization of the perceptual generator with input from the senses.

Inspired by Stephen Thaler’s (1996a, 1996b) Crea-tivity Machine Paradigm, we imagine the perceptual and cognitive generators as interacting information networks in which sensory input parameterization to the percep-tual generator is optional. We can visualize these ideas (Figure 2) by metaphorically depicting sensory input, and the perceptual and cognitive generators as gears: a “sensory input gear,” a “perceptual gear” (representing the perceptual generator and its conscious output), and a

Sensory Input

Perceptual Environment

Generator

Perceiving “I”;Cognitive Generator

CASE 1: WAKING CASE 2: DREAMING

Sensory Input

Perceptual Environment

Generator

Perceiving “I”;Cognitive Generator

Fig. 2. Simplified view of conscious processes during waking and dreaming. During waking, sensory input serves to parameterize the generation of a perceptual environment; the consistent perceptual structure serves as a basis for consistent cognitive activities. During dreaming, loss of sensory input parameterization leads to inconsis-tency in the perceptual generator, and increased lability between perceptual and cognitive activities of consciousness.

Global Workspace and Dreaming 7

“cognitive gear” (representing the cognitive generator and its conscious output). During waking, the three gears engage each other and move in synchrony. Sen-sory input gives content to, or “parameterizes,” the per-ceptual generator. The self, encompassing the conative, cognitive and metacognitive components of conscious-ness, reacts to and acts upon its perceived environment in a consistent fashion. However, during sleep, sensory thresholds increase across all sleep stages (Bonnet, 1986; Bonnet & Johnson, 1978; Busby, Mercier, & Pivik, 1994; Niiyama, Sekine, Fushimi, & Hishikawa, 1997; Rechtschaffen, Hauri, & Zeitlin, 1966), which can be construed as “disengaging” the “sensory gear” from the equation, and leaving only the “perceptual” and “cognitive” gears in functional operation. The operation of the cognitive and perceptual generators minus sen-sory input parameterization results in the conscious sleep experiences described above. We now fill in the details of this model by first focusing on waking experi-ence, and then returning to a more detailed discussion of nonlucid and lucid dreaming.

OVERVIEW OF THE GLOBAL WORKSPACE SYS-TEM DURING WAKING

When we talk about the perceptual and cognitive generators above, we are discussing the major, salient, phenomenological aspects of conscious awareness. Our normal waking subjectivity contains consciousness of an external perceptual environment and of being an “I” with memory access, volition and other cognitive abili-ties within this environment. Both aspects of con-sciousness have substructure and both ride astride many complex, some as yet undetermined, unconscious proc-esses. We now refine and expand upon these concepts within the framework of Bernard Baars’ GW model.

The GW model (Baars, 1988; Baars, 1997a; Neu-man & Baars, 1993) is an elegantly simple, yet powerful framework that envisions waking psychology as an un-ceasing interplay of conscious and unconscious proc-esses (Figure 3). Conscious processes manifest as a “global workspace” which in turn is structured and con-ditioned by a host of unconscious neurocognitive proc-esses. The central metaphor of the GW model is quite simple: consciousness serves as a system-wide receiving and broadcast medium for the entire nervous system. To use Baars’ terminology, consciousness is the “publicity organ” of the nervous system. Unconscious elements input structure into consciousness such as perceptions, emotions, thoughts, and motivations. Consciousness in

turn serves to broadcast its contents to the entire nervous system, and in doing so activates or recruits other rele-vant unconscious processes related to attentional, cogni-tive, metacognitive, mnemonic or effector actions. In this sense then, consciousness is a global “workspace” - a global medium - in which transactions of information occur throughout the entire CNS.

CONSCIOUSNESS DURING WAKING

The “stream of consciousness” is clearly a compos-ite of many facets including perceptions, imaginings, recollections, attention, emotions, thoughts, metacogni-tions and volitions. Baars suggests that conscious opera-tions can be divided into three areas. Conscious ex-periences are perceptual or “quasi-perceptual” and are the direct, immediate content of our consciousness: sen-sory-based perceptions, internal imagery, emotions, thoughts, recollections and so forth. Consciously-mediated access involves access to contents that are not currently conscious, but can be made so readily. This suggests a type of “penumbra” surrounding focal consciousness, an area of “preconsciousness” to which consciousness can immediately shift (Baars, 1996a; La-Berge & Rheingold, 1990). Conscious access opera-tions are closely associated with voluntary attention, voluntary memory access, and imagery operations (Baars, 1997b). Conscious-mediated control is, es-sentially, volition, the ability to consciously initiate ac-cess and action operations.

We can map Baars’ three aspects of conscious op-erations onto our dichotomy of perceptual and cognitive generators, as shown in Figure 4, illustrating the gross structure of the global workspace of consciousness as we will define it in this paper. Conscious control and access operations are initiated by the self, and hence considered by us as products of the cognitive generator. Conscious experiences, however, are produced by both the cognitive and perceptual generators. Consciousness experiences of an external environment are the most vivid aspects of conscious experience (Baars, 1988). However, conscious experiences of internal imagery, including modal (visual, auditory, verbal, tactile, kines-thetic, etc.) and amodal (semantic, intuitive) imagery, and affect, are initiated by and/or occur within the self, and are therefore products of the cognitive generator within our framework.

8 DeGracia

Such an assignment of conscious experiences across both generators may seem somewhat arbitrary but for the following considerations: (1) the perceptual genera-tor is directly parameterized by sensory input, but the cognitive generator is not, which is why, during both waking (Kosslyn, 1994) and dreaming (Worsley, 1988), the imaginal contents of consciousness can occur in par-allel to the perception of an external environment. (2) Neuroimaging data indicates that modality specific in-ternal imagery shares components with networks medi-ating direct perceptual experiences (Chen, Kato, Zhu, Ogawa, Tank, & Ugurbil, 1998; Farah, Peronnet,

Gonon, & Girard, 1988; Kosyln, 1994). However, im-agery and perception are two different systems both anatomically (Howard, Fytche, Barnes, McKeefry, Ha, Woodruff, Bullmore, Simmons, Williams, David & Brammer, 1998) and functionally (Mellet, Petit, Ma-zoyer, Denis, & Tzourio, 1998), and the fidelity of these systems can be compromised by dividing attention be-tween them (Baars, 1988; Posner, 1982). (3) The dis-tinction between perceiving what appears to be external to our self and internal to our self is consistent with the phenomenology of normal subjective experience during both waking and dreaming.

GW

unconscious processors

Baars (1988) offers further characteristics of con-scious processes. Conscious processes are computa-tionally inefficient in some respects; they are error prone, slow, of limited capacity, subject to mutual inter-ference, and are serially ordered over time. On the other hand, consciousness provides a medium for associating an almost endless variety of contents (a relational capa-bility), but is also acutely sensitive to context, which provides for an internal consistency to conscious con-tents. Our precise usage of the term “context” will be provided below.

Fig. 3. The GW model portrays consciousness as a “global work-space” (GW) supported by unconscious processors that input struc-ture into consciousness and use the contents of consciousness to mediate changes in the nervous system. Diagram modeled after Baars (1988).

UNCONSCIOUS PROCESSES DURING WAKING

In contrast to conscious processes, Baars (1988) points out that unconscious processes are limited in functional range (e.g. are dedicated to one or a few processing functions) but, taken as a whole, have a large computational capacity. Unconscious processes are computationally more efficient; they occur in parallel, have a low error rate, high speed and display little mu-tual interference. The detailed operation of these special-

Conscious processes Unconscious processes Computationally inefficient; High number of errors, low speed, and mutual interference between conscious computations.

Highly efficient in their own task; Low number of errors, high speed, and little mutual interference

Great range of different contents over time (differentiation and com-plexity); great ability to relate different conscious contents to each other (relational capacity);great ability to relate conscious events to their unconscious contexts (context-sensitivity); conscious contents are informative.

Each specialized processor has limited range over time; each one is relatively isolated and autonomous

Have internal consistency, seriality, and limited capacity. Diverse, can operate in parallel, and together have great capacity.

Table 1: Comparison of conscious and unconscious processes. Adapted from Baars (1988).

Global Workspace and Dreaming 9

ized subsystems critically affects the contents of con-sciousness yet remains outside of direct conscious awareness. A comparison of conscious and unconscious processes is provided in Table 1. In bringing the notion of unconscious processes explicitly into our discussion we begin to introduce important details to the models presented in Figures 2 and 4. More specifically, con-scious experience, access, and control operations are mediated by unconscious processes (Baars, 1988; Rock, 1997).

Two examples illustrate the distinction between the terms “conscious” and “unconscious” as used in the GW model. Opening one’s eyes results in the intense con-scious experience of seeing. Seeing is not merely pas-sive; directed attention, motivation, and goals influence what we see (reviewed in Theeuwes, 1994). However, beyond these modulating influences, the generation of vision proceeds automatically and unconsciously. We know a great deal about unconscious visual processing modules and their pathways, from the actions of light in the retina (Sterling, 1998), to the complex columnar structure of area 17, the generation of color and complex form in the ventral visual pathway, and the generation of motion, depth and spatial perception in the dorsal visual pathway (reviews can be found in Gulyas, 1997; Schiller, 1997; Zeki, 1997). The activity of these sys-

tems generates our conscious experience of seeing, and yet occurs outside of direct conscious awareness, which is presumably the composite end result of these lower level unconscious processing stages (Marr, 1983; Moore & Engel, 1999). A second apt example of the role of unconscious processing is the reading of these words. As you read these words, you quite automatically and unconsciously convert patterns of light and dark contrast to meanings in your consciousness. There are many intervening neurocomputational stages of visual and linguistic processing between seeing this sentence and knowing the meaning of these words in your mind. The intervening visual, syntactical, grammatical, semantic, and mnemonic processing stages are essentially invisible to your consciousness (Baars, 1988). And, wth a lttl mr cnscs effrt u cn qte atmtclly dcd ths sntnc. The forego-ing is a feat of pattern recognition far beyond any cur-rent technology, and its neurocomputational underpin-nings happened unconsciously, automatically.

Cognitive Generator (“I”)

Perceptual Generator (not “I”)

Conscious control

Conscious access

Internalized conscious experiences

Externalized conscious experiences

Fig. 4. The Global Workspace of consciousness. Mapping the ideas of cognitive and perceptual generators to Baars’ (1988) functions of conscious control, access, and experience. Conscious control and access are mediated by the “I” of the cognitive generator. Conscious experiences are a product of both the perceptual and cognitive gen-erators; the former producing externalized objects of awareness, the latter producing both modal and amodal forms of internal imagery and recall.

More precisely, Baars’ use of the concept “uncon-scious processors” corresponds almost exactly to Llinás and Paré’s (1996) concept of the “neurological a priori” properties of the brain. To use examples provided by these latter authors, abilities such as hearing, seeing, acquiring language, and so forth do not have to be learned; these are genetically determined properties of the human brain, although their optimum expression is use-dependent (Black & Greenough, 1986). Both of these concepts are widely recognized today as referring to the discreet “processing modules” associated with specific cortical and subcortical regions.

On the basis of the above considerations, we can expand our conception of the perceptual and cognitive generators to indicate that they are generating conscious operations from a host of unconscious processors (Fig-ure 5). In fact, we can now state that inclusion of sen-sory parameterization in Figure 2 was but one specific type of unconscious influence on conscious processing. Sensory input is unique, however, because it is func-tionally “optional”; it dominates waking, but is attenu-ated during sleep. In Figure 5, the generators are now depicted as boxes with a third dimension of depth. The top surface of each box indicates the conscious output of the generators, the depth is meant to convey the uncon-scious processes occurring underneath the surface of consciousness, “behind the scenes” so to speak.

10 DeGracia

Perceptual Generator (not “self”)

1. Visual mappings/schemata2. Somatopic mappings/schemata a. pain b. temperature c. somesthesia d. kinesthesia3. Auditory mappings/schemata4. Olfactory mappings/schemata5. Gustatory mappings/schemata6. Viscerotopic mappings/schemata7. Unimodal to multimodal transformations

UNCONSCIOUS PERCEPTUALPROCESSORS

Conscious experiences of external “reality”

Cognitive Generator (“self”)

1. Affect/Emotion/Motivation2. Memory systems a. episodic memory b. semantic memory c. action/skill memory d. working memory3. Attentional systems4. Metacognition/Executive functions5. Effector control6. Linguistic processors: a. phonetic b. syntax/grammar c. vocabulary

UNCONSCIOUS COGNITIVEPROCESSORS

Conscious experiences ofinternal events; consciousaccess and control

Fig. 5. The conscious output of the perceptual and cognitive generators is mediated by a host of unconscious neurocognitive processes.

The unconscious processors of the perceptual gen-erator involve brain regions dedicated to sensory proc-essing. Anatomically, these consist of complex feedback circuits between unimodal and multimodal sensory cor-tex and their respective thalamic nuclei (the thalamocor-tical loops of Llinás and Paré, 1991), with some varia-tion in pain and olfactory systems (Smythies, 1997). During waking, perceptual modalities are parameterized by input from ascending sensory pathways which, in fact, represent a minority of input to the perceptual gen-erator (Erisir, Van Horn, & Sherman, 1997; Sherman & Koch, 1986; Wilson, Friedlander, & Sherman, 1984). In Figure 5, the term “mappings/schemata” used to de-scribe the unconscious processors of the perceptual gen-erator refers to the sensory-specific representations of each modality, and the multiple topographic mappings and coordinate transformations that exist throughout cortical and subcortical sensory pathways (Cline, 1998; Kaas, 1997; Obermayer, Sejnowski, & Blasdel, 1995). Unimodal sensory cortices contain relatively discreet processing modules capable of representing different aspects of a given modality (Felleman & Van Essen, 1991; Mesulam, 1998). Unimodal sensory pathways converge to multimodal cortices where evidence sug-gests that alignment of various unimodal mappings are transformed into multimodal representations (Bertenthal, 1996).

The cognitive generator is clearly more complex

than the perceptual generator. Our list of unconscious processors mediating conscious cognition of the self (Figure 5) is not meant to be exhaustive, but representa-tive of such processors, all of which are subject to con-scious access and control by the self. These include fa-cilities for mediating directed attention, various forms of voluntarily accessible memory (including declarative and short-term/working memory), affect/emotion, moti-vation, forms of executive control (D’esposito & Grossman, 1996) such as volition and metacognition (thinking that reflects on thinking), and effector control.

The neuroanatomical substrates of cognition are not fully understood presently, but do involve higher order cerebral cortices and their associated subcortical connec-tions. Multimodal association cortices project to supra-modal cortical sites, where highly processed sensory information converges with limbic, attentional, motor, mnemonic, and other forms of nonsensory information (Cohen & Eichenbaum, 1993). Together these systems serve as nodes in parallel distributed networks (Bullmor, Rabe-Hesketh, Morris, Williams, Gregory, Gray, & Brammer, 1996; Felleman & Van Essen, 1991; Gold-man-Rakic, 1995), from which cognitive functions arise. Since our main concern in this paper is developing a framework for discussing the general structure of sub-jective consciousness, neuroanatomical loci or neuro-physiological functions associated with aspects of cog-nition will be introduced as needed in the discussion.

Global Workspace and Dreaming 11

Clearly, perception structures cognitive functions such as thinking and planning. Likewise, cognitive op-erations direct the machinery of perception. Thus, an important aspect of perceptual and cognitive generator function is that they interact with one another. By “in-teraction” we mean the transfer of information between the perceptual and cognitive generators. An important distinction of generator interactions is whether these occur in consciousness or via unconscious forms of processing.

Figure 6 provides a graphical depiction of the ideas stated to this point. Consciousness (the GW) is di-chotomized into representations of an external percep-tual environment and the conscious cognition associated with the self. Both of these derive from the concerted action of many unconscious processing modules, which are depicted simply as the perceptual and cognitive gen-erators. Figure 6 also shows the information flow path-ways possible in this model. Information flows into the GW from the unconscious processors and in turn, the GW can project information back to the unconscious processors in both generators. That is, information can be indirectly transferred between generators using con-sciousness as an intermediary. Information can also be directly transferred between the modules of the genera-tors, bypassing conscious awareness altogether.

COOPERATION, CONTEXTS, AND PERSONALITY

We now elaborate in some detail on how the infor-mation content of unconscious modules affects con-scious operations by describing Baars’ (1988) concepts of “cooperation” and “contexts,” ideas intimately related

to learning and memory. These ideas are central in our application of the GW model to dreaming. Global

WorkspaceCC

PE

UnconsciousProcessorsCG PG

Fig. 6. Our adaption of the GW model dichotomizes consciousness into conscious perception of the external environment (PE = percep-tual environment) and the conative aspects of conscious cognition (CC). This diagram illustrates potential pathways of information flow operating at either conscious or unconscious levels of neuro-cognitive processing. See text for details.

“Cooperation” means that transient coalitions of un-conscious processes can be recruited on an “as needed” basis and thereby affect conscious operations as a unit. Such a view is consistent with the results of neuroimag-ing studies showing activation of different networks of brain regions during execution of specific cognitive and behavioral tasks (Posner, DiGirolamo, & Fernandez-Duque, 1997; Posner & Raichle, 1994). A similar type of modular cooperation is to be found in cognitive mod-els of linguistic production in which discreet processing modules for phonetics, grammar, syntax, semantics, and pragmatics are hypothesized to interact (cooperate) in the flow of language comprehension and production (Foss & Hakes, 1978; Winograd, 1972).

However, regularities in perceptual and cognitive activities produce regularities in cooperative activity among unconscious processors. These more permanent associations of unconscious processes Baars’ terms “contexts.” Baars defines contexts as “relatively endur-ing [information] structures that are unconscious, but that can evoke and be evoked by conscious events” (Baars, 1988, p. xx), and also as “a system (or set of sys-tems) that constrains conscious contents without itself being conscious” (op. cit., p. 372).

To provide a bridge between Baars’ psychological concept of “context” and current ideas in the neurobiol-ogy of memory, we now introduce a formal model of memory presented in Cohen and Eichenbaum (1993). The intent of this model is to explain results from mem-ory studies of normal and amnesic human subjects, and lesion data from animal studies. The Cohen and Ei-chenbaum model defines two forms of memory in the CNS: declarative and procedural memory. The authors’ own words best describe this distinction:

“…the declarative memory system receives, and plays

a mediating role in the storage of, the outcomes of proc-essing events. Declarative memory is a fundamentally re-lational representation system. The relational nature of de-clarative memory gives rise to two proper-ties…representational flexibility and promiscu-ity…declarative memory is promiscuously accessible to, or can be activated by, various processing modules, re-gardless of which processing modules were engaged in the processing of the original learning event; and, once ac-cessed, it can be manipulated and flexibly expressed in various…contexts, regardless of how much those contexts differ from the circumstances in which the information

12 DeGracia

was initially acquired.” (Cohen & Eichenbaum, 1993, pg. 73)

“By contrast, procedural representations, supported by memory systems that operate independently of the hip-pocampal system, are inflexible and dedicated. Their storage resides with and remains inextricably linked to the processing modules that were engaged during the initial learning. This type of memory involves not the storage of outcomes of processing operations, but rather tuning of and changes in the way those operations actually run--that is, modifications of the processing elements themselves. This type of representation is therefore inflexible; it is only accessible to those processing modules that were en-gaged during the original learning experiences, and only when they are again engaged. The representations, there-fore can only be expressed or otherwise exert their influ-ence under...conditions that so closely mirror the ..original learning...as to constitute a repetition of the original learn-ing situation.” (Ibid, pg. 74).

To put the above description of the declarative

memory system in perspective, we need some sense of input/output relationship of this system with other brain regions. Therefore, we here introduce some of the anat-omy of this system, which will be important later when we present a neurobiological model of the formation of the dream environment.

Declarative memory is mediated by what we shall term in this paper the medial temporal lobe memory sys-tem (MTMS). The MTMS includes the hippocampal system (HS = hippocampus proper, subicular com-plexes, and dentate gyrus [DG]), and related rhinal cor-tices (the entorhinal [EC], parahippocampal [PHC], and perirhinal [PRC] cortices) (Cohen & Eichenbaum, 1993; Suzuki, 1996). The information flow into the MTMS involves highly processed outputs from higher order cortices feeding into the PHC and PRC. These project to the EC (Amaral, Insausti, & Cowan, 1987; Braak & Braak, 1992; Insausti, Amaral, & Cowan, 1987) which in turn projects to the HS. All of these pathways are bidirectional so that HS output follows the reverse pathway. A highly oversimplified flow diagram is:

higher order unimodal and supramodal cortices

PHC & PRC EC HS Although there is significant overlap, the input to

the rhinal cortices displays a gross modality specificity. The PRC receives a large input from visual pathways in the inferior temporal lobe related to object recognition, form and color processing (e.g. visual areas TE and

TEO) (Suzuki & Amaral, 1994a). The PHC is further divided into areas TH and TF (Bonin & Baily, 1947). TH has significant auditory and spatial inputs (Suzuki & Amaral, 1994a). Area TF has strong somatosensory and visual-spatial inputs, including connections with dorso-lateral prefrontal cortex and posterior parietal lobe (Cavada & Goldman-Rakic, 1989). The PHC and PRC share significant bidirectional interaction amongst each other and with the EC (Suzuki & Amaral, 1994b). All of the rhinal cortices are further bidirectionally con-nected with multi- or supra-modal sites in the prefrontal cortex, dorsal superior temporal sulcus, cingulate cortex or retrosplenial cortex (Suzuki & Amaral, 1994a; Su-zuki, 1996; Witter, Groenewegan, Lopes da Silva & Lohman, 1989). Thus, there is a huge convergence of highly processed cerebral information serving as input to the MTMS; the outcome of sensory processing con-verges with information pertaining to attention, affect, motivation, planning, and motor behavior (Eichenbaum & Otto, 1993).

Cohen and Eichenbaum discuss why there would be two types of memory systems in the CNS. The declara-tive system is a set of processing modules whose func-tion is to flexibly relate the output from other processing modules into long-term declarative memories, which then become promiscuously accessible to the entire nervous system. Procedural memory, on the other hand, seems to be a generic property of all of the processing modules, such that repetitive activity in a given module will fine-tune and optimize the operating characteristics of that module. Hence, procedural memory is a local form of memory apparently innate in nervous tissue, whereas declarative memory is a global form of mem-ory, requiring specific modules (the MTMS) to instanti-ate it. Cohen and Eichenbaum recognize that the de-clarative system is in fact a multiple memory system from the point of view of content, and that legitimate distinctions can be drawn for episodic, semantic, or other content-specific forms of declarative memory (Schacter & Tulving, 1994; Squire, 1994; Tulving, 1987).

Cohen and Eichenbaum’s declarative/procedural distinction is strikingly similar to Baars’ functional dif-ferentiation between conscious and unconscious proc-essing. Consciousness is associated with global CNS access, content flexibility and relational capability. Un-conscious processors are associated with local, dedi-cated processing. Clearly, declarative memory is an im-portant component underlying conscious processing op-

Global Workspace and Dreaming 13

erations, and procedural memory underlies changes in unconscious processing modules. On this basis, it is not unreasonable to assert that the contents of consciousness serve as input to the declarative memory system; not all conscious contents enter the declarative system, but all input to the declarative system comes from conscious contents. As Baars (1988) has pointed out, associative learning only occurs in conscious animals: even Pav-lovian conditioning cannot be performed on an uncon-scious animal.

Baars’ concept of “context” encompasses both de-clarative and procedural memory systems. When changes occur in a single unconscious processor, leading to changes in conscious operations, this is a form of con-textual learning in the GW system. Examples of single processor changes affecting consciousness, without the change being conscious, might include perceptual (Schacter & Buckner, 1998) or conceptual priming (Graf & Schacter, 1985), or the learning of motor skills (Wil-lingham, 1998), activities which would improve reaction times in repetitions of the original learning situation.

However, patterns of relationships amongst proces-sors could lead to higher order procedural modification of many separate processors simultaneously. This is an implication not explored in Cohen and Eichenbaum (1993); that the declarative memory system itself could be subjected to procedural memory effects. Which is to say, styles of learning and styles of relating information entering the declarative system can become optimized and fine-tuned over time. The procedural fine-tuning of the declarative memory system would occur with expo-sure to repeating patterns of learning and relationship encountered during perceptual or behavioral activities. Hence, within the declarative/procedural memory dis-tinction, Baars’ concept of “context” actually encom-passes two distinct levels of the organization of mne-monic information in the nervous system: (1) it can refer to procedural modifications of individual processors, or more importantly, (2) it can refer to higher order pat-terns of relationship produced by the declarative system but made implicit over time by procedural effects within the MTMS. In this paper, our use of the term context will be meant to refer to the second usage, unless other-wise specified.

It is this second sense of “context” by which Baars (1988) describes how the specific information content of the CNS forms higher order memory structures which impart an implicit “framing” effect on conscious proc-esses. Baars defines specific types of contexts, which are

relationships between multiple processors that are im-plicit, not explicit, in on-going conscious operations. For example, he defines “goal contexts” as complex mnemonic composites, combining semantic and epi-sodic memory, motivational, and effector modules re-quired to envision and execute specific goals. During the real-time execution of a goal, the overall goal-context fades from direct consciousness, but continues to implicitly frame real-time conscious activities. Baars defines “belief contexts”, which can be thought of as patterns of learning by which the individual conceptual-izes and organizes new experiences. Over time, these patterns of learning become progressively more implicit in, yet still exert influence over, conscious activities. In section 6.1.2 we shall define a specific multicomponent mnemonic structure, the “lucid dream context,” the op-eration of which, we will argue, distinguishes lucid from nonlucid dreams.

Returning to our dichotomy of the perceptual and cognitive generators, the modules of both of these are subject to procedural and declarative effects, and can participate in the formation of higher order contexts. Perceptual contexts will frame conscious perceptions of the external environment. A simple example is that when we look outside, the sky is always up. It is not unreasonable to imagine that, through constant repeti-tion of this visual stimulus, visual-spatial brain regions are highly optimized at always representing the sky in an upward direction relative, say, to the overall visual scene or the pull of gravity. Such a “the sky is always up” context would explain why the sky is typically in an upward direction in dreams. Similarly, habits of think-ing, habits of learning, habits of self-perception, etc. will form contexts within the cognitive generator.

Baars, however, goes even further and suggests that different contexts do not exist in isolation from one an-other, but form “context hierarchies,” where broader, more encompassing contexts contain within them more local contexts. This idea culminates in the notion of a “Dominant Context Hierarchy” which Baars equates with representations of Self. That is, the Dominant Con-text Hierarchy is the sum total of all contextual organi-zations in an individual’s nervous system. At this level of conceptualization, in fact, the GW model is a model of the structure of the human personality and, as such, is highly consistent with cognitive-affective personality paradigms in personality research (Mischel & Shoda, 1998) and other models of the self (Damasio & Dama-sio, 1996).

14 DeGracia

The impetus behind these ideas is clearly to under-stand how information is organized and utilized in the nervous system, and specifically how preexisting memo-ries provide implicit organizations which frame, struc-ture, or limit real-time conscious operations. We feel it is critical to recognize that Baars’ notion of Dominant Context Hierarchy converges to the psychology of per-sonality. When we discuss changes in conscious and unconscious processes during dreams, we will take the waking personality, conceptualized as a structure of con-textual hierarchies, to be our basic unit for all across-state comparisons.

COMPETITION AMONG UNCONSCIOUS PROC-ESSES

The contextual structuring of conscious operations is a neurocomputational strategy for dealing with the limited capacity of conscious operations; with the for-mation of habits of thinking and action, any given con-scious operation contains more information than it would otherwise. Another method, elaborated by Baars (1988, 1993), for increasing the efficiency of conscious operations, is to set barriers for access to consciousness by forcing unconscious processes to compete with each other for access to consciousness. One conscious result of unconscious competitive processes involves the shift-ing of attention; when unconscious processes compete, the “winner” gains access to consciousness (dominates the global workspace, to use Baars’ terminology) and typically becomes the focus of attention. In this fashion the salient and most urgent information gains access to and control of consciousness.

Baars (1988) defines two forms of competition: (1) competition between potentially conscious stimuli, and (2) competition between different context hierarchies. The former clearly is relevant to the perceptual genera-tor, the second form of competition occurs within and/or among both generators. In the first, and better studied case, competition can occur between sensory modalities, or between stimuli within a modality, directly affecting conscious perceptual experiences (Flowers, 1990). An example of perceptual competition is the Necker cube, which can only be seen in one perspective in any instant. Competition amongst different context hierarchies in-volves significant cognitive complexity. A simple ex-ample may be whether one should go golfing or write the grant whose deadline approaches. Such competition involves displacement of the current contextual hierar-chy by another, altering the implicit structuring of con-

scious operations (Baars, 1988). This would manifest consciously as shifts in mental set and orientation, shifts in attitude, or shifts in goal objectives. Competition for conscious experience can also occur between the per-ceptual and cognitive generators, for example, the com-petition between perception and imagery (e.g. Green-berg, 1977). We utilize the idea of competition between context hierarchies in section 7.2 to explain the observa-tion that lucid dreams are relatively susceptible to rever-sion to nonlucid dreams.

NOVELTY, LEARNING, AND CONSCIOUSNESS DUR-ING WAKING

The GW model contains important functional impli-cations for waking conscious operations. The transient cooperation of unconscious processors imparts a large degree of flexibility to conscious operations. Waking consciousness can be seen as a mechanism for respond-ing to novelty, and for providing a medium of associa-tion between a wide variety of potentially conscious contents. On the other hand, regularities of perception and action produce regularities in cooperative activity among processing modules, leading to the contextual structuring of conscious operations. In this regard, con-sciousness can be seen as the medium through which repetitive forms of perception, cognition, and action are converted into the more efficient form of relatively automatic and unconscious habits and routines via both procedural and declarative memory systems. Thus, waking consciousness is a (global) “work space” sub-stantially linked to adaptability and learning; conscious contents can reorganize into new patterns, but can also transfer relevant patterns to long-term storage as more efficient unconscious contexts. Clearly, as psychological maturation proceeds, increasingly established perceptual and cognitive contextual structures will limit the flexi-bility of waking conscious operations. As dream theo-rist have noted, dreaming consciousness is characterized by a “hyper-associativity” (Freud, 1900; Hobson, 1988; Kahn et al., 1997). In section 9 we will formulate this observation in terms of the flexibility aspect of con-sciousness strongly predominating over the formation of habit-responses during dreams, which will allow us to suggest a function for dreaming within the overall econ-omy of the nervous system.

Global Workspace and Dreaming 15

FUNCTIONAL REORGANIZATION OF THE BRAIN DURING SLEEP

There is divergence among dream theorists as to the role of sleep neurobiology as a causative factor in sleep psychology. Some investigators question a direct causal link between sleep biology and psychology (Foulkes & Cavallero, 1993; Moffitt & Hoffmann, 1987). Other in-vestigators have constrained their views of sleep psy-chology based on our current knowledge of sleep neuro-biology (Hartmann, 1982; Hobson & McCarley, 1977; Kahn et al., 1997; Koukkou & Lehmann, 1983). Clearly, the brain undergoes significant changes during sleep, and it would be unreasonable to believe that this cannot directly alter the psychological functioning, and even gross function, of the sleeping brain. We now out-line the salient aspects of sleep neurobiology to set a stage for discussing the gross psychology of sleep.

The most obvious feature of the sleeping brain is the sleep cycle: the time course of alternation between stages I-IV NREM and REM (Rechtschaffen & Kales, 1968). Considerable evidence from animal-based stud-ies shows that these periodic alterations are mediated by changes in brainstem neurotransmission involving pri-marily the serotonergic raphe nuclei, noradrenergic lo-cus ceruleus, and the cholinergic pedunculopontine and laterodorsal tegmental nuclei (reviewed in Gaillard, 1985; Hobson & Steriade, 1986; Jones, 1991). The re-ciprocal interaction model (Hobson & Schmajuk, 1988) states that aminergic transmission decreases while cho-linergic transmission increases across NREM stages I-IV, reaching their minimum and maximum, respectively, during REM sleep. The original activation-synthesis model posited that cholinergic transmission served as a diffuse activation source for the forebrain via PGO waves; this view has been expanded by culminating evi-dence that there are changes in other neurotransmitter systems (reviewed in Kahn et al., 1997). One of the more well-established consequences of altered neuro-transmission during sleep is an attenuation of sensory-input processing (Geof, Allison, Shapiro, & Rosner, 1966; Velasco, Velasco, Cepeda, & Munoz, 1980). Considerable evidence indicates that changes in brain-stem neurotransmission during sleep are consistent with increased sensory thresholds and alternations of burst- and transmission-modes in the thalamo-cortical percep-tual system (reviewed in Guido & Lu, 1995; McCor-mick, 1992; Steriade & Llinás, 1988).

Ideas of a diffuse activation of the brain during sleep, or ideas of asymmetrical cerebral activation (An-

trobus, 1987; Green & McCreery, 1994) have had to be modified by evidence that patterns of forebrain activity are more focal than had initially been suggested by EEG studies. Two lines of investigation have provided evi-dence for focal changes in brain activity during sleep in humans: neuroimaging studies, and neuropsychological studies of the dreams of patients with focalized brain lesions.

The neuroimaging studies, in particular, are in their earliest stages and have produced both conflicting and consistent results. For example, three imaging studies of REM all saw increased activity (after subtraction of waking scans) in the amygdala, anterior cingulate gyrus and temporal lobe regions associated with the MTMS (e.g. parahippocampal, and/or entorhinal cortex) and all saw decreased activity in the posterior cingulate gyrus (Braun, et al., 1997; Maquet, Peters, Aerts, Delfiore, Degueldre, Luxen, & Franck, 1996; Nofzinger, Mintun, Wiseman, Kupfer, & Moore, 1997). However, each study reported activity changes in regions not necessar-ily reported by the other two studies. For example, Braun et al. (1997) and Nofzinger et al. (1997) saw in-creased basal ganglia activity, which was not reported by Maquet et al. (1996). Descriptions of changes in cerebral cortical regions were particularly disparate be-tween the three studies. These discrepancies are likely due to differences in imaging procedures, sleep para-digms, subject samples, and what was going on in the minds of the subjects at the time of brain scans.

Data about sleep psychology acquired from brain damaged patients is difficult to interpret. Patients with similar symptomology will be grouped together, but none of the patients have exactly the same pattern of brain damage. Brain damage interrupts white matter tracts as well as lesioning nuclei, making functional as-signments ambiguous (Farah, 1994). Furthermore, ob-taining such data is difficult: these patients often cannot talk or comprehend language well, and have sensory, motor, mnemonic, emotional or motivational deficits to varying extents. Distinguishing between dream recall and true cessation of dreaming is problematic, although not insurmountable (Solms, 1997). Thus, trying to ob-tain data about sleep experience in this way is a brave task. However, some generalizations have emerged. Specific types of focal damage result in alterations in dream consciousness or in global cessation of dreaming (Epstein, 1979; Murri, Arena, Siciliano, Mazzotta, & Muratorio, 1984; Schanfald, Pearlman, & Greenberg, 1985; Solms, 1997). For example, occipital damage

16 DeGracia

producing visual irreminiscence during waking leads to loss of visual dreaming (Solms, 1997). Mark Solms (1995, 1997), in perhaps the most extensive study of the dreams of brain lesioned patients to date, has suggested that damage to two rather gross regions - the white mat-ter tracts feeding the ventral frontal lobes, and the poste-rior parietal cortex - can lead to complete cessation of dreaming. We return to Solms’ thinking below.

Although Hobson, et al. (1998b) has claimed that the neuroimaging studies are consistent with the patient data, it is clear from our brief review that regions consis-tently displaying high activity in neuroimaging studies are not the same regions Solms claims are necessary for preventing cessation of dreaming. Perhaps the largest inconsistency in this regard is that Solms’ patient data indicates the continuation of dreaming in patients with pontine damage, ostensibly contradicting the central role of brainstem transmission suggested by animal-based studies, and at least one of the neuroimaging studies (Braun et al., 1997). Clearly then, increased regional activity does not imply that activity is necessary and sufficient for dream production.

Given the limits of present imaging technology and neuropsychological analysis, we have gained intriguing hints that suggest definite patterns of cerebral activation, but nothing so definite we can build unassailable models of dreaming solely on these bases. In section 8.3, we utilize the above data, in conjunction with phenomenol-ogical data of dream subjectivity, to construct a neuro-biological model of the genesis of the dream environ-ment. However, we present our model clearly recogniz-ing its susceptibility to revision as new and more precise data emerges. What is clear presently is that the brain undergoes substantial functional changes during sleep which represent clear discontinuities across the sleep-wake cycle. We believe a fusion of continuity and dis-continuity views is required. From a closed system per-spective of the CNS, there is continuity because it is fundamentally the same system operating across the sleep-wake cycle. The discontinuities, evident in both the biology and the psychology (as we discuss below), result from a functional reorganization of the brain-mind system during sleep.

DREAM CONSCIOUSNESS AS ACTIVATION OF THE COGNITIVE AND PERCEPTUAL GENERATORS

Hunt (1989) explicates the subtle difficulties of de-fining dreams and of comparing them to waking experi-ence. Perhaps counterintuitively, distinguishing dream-

ing from waking in a fashion all can agree upon has proven elusive. Several researchers analyzing the spe-cific content of dreams have been impressed with the general similarity between dreams and waking experi-ence (Boss, 1958; Dorus, Dorus, & Rechtschaffen, 1971; Hall & van de Castle, 1966; Hunt, Ogilvie, Be-licki, Beliki, & Atalik, 1982; Kahan, et al., 1997). Other investigators stress differences between waking and dreaming (Antrobus, 1987; Foulkes, 1991; Hobson, 1988; Retschaffen, 1978).

As we stated in section 2, the phenomenology of subjective experience during sleep indicates that there is a spectrum of activation operating in both the perceptual and cognitive domains. The idea of “activation” figures prominently in several conceptions of dreaming. The original formulation of the activation-synthesis model was that ascending brainstem impulses serve as a source of forebrain activation (Hobson & McCarley, 1977), a necessary prerequisite to produce the rich conscious content of dreams (Hobson & Steriade, 1986). The AIM model has formalized the continuous variation of cere-bral activation during sleep (Hobson & Stickgold, 1995). John Antrobus’ notion of “cortical activation” during dreaming refers to the spread of activation through network nodes (Antrobus, 1986; Antrobus, 1991). Hunt’s (1989) notion of “intensity spectrum” de-scribes the range of cognitive fidelity occurring in dreams: confusion and clouding of consciousness at one extreme, ego-transcending transpersonal states of con-sciousness at the other extreme of this spectrum. Le-querica (1996) has also explicitly recognized a contin-uum in dream cognition. With the exception, perhaps, of the original formulation of the activation-synthesis model, what all of these views share is the recognition that cognition in dreams occurs along a continuum, be it cortical activation, fidelity of cortical networks, or de-grees of cognitive expression.

What all of these views lack is the recognition that the dreamer and dream environment can be conceptual-ized as two separate entities that can undergo alterations seemingly independent of one another. From a third person perspective (e.g. reading dream reports by oth-ers), it is natural and logical to think of the dream per-ceptual environment as just another element within the dreamer’s consciousness, and lump changes in the dream environment along other aspects of the dreamer’s cognition. Again, however, from the first person view-point we have experienced many times in the dream state, it is clear to us that the dream environment seems

Global Workspace and Dreaming 17

to behave as if it is a thing quite independent of our con-scious cognitive activities. As well, we have both per-sonally observed variations in our own conscious cogni-tion during dreaming, being at times more like waking and at times less than waking, and these cognitive altera-tions appear to occur independent of the dream percep-tual environment. Hence our dual characterization.

We explicitly associate the dreamer and the dream perceptual environment with the activities of the cogni-tive and perceptual generators, respectively, as these were defined throughout section 4. Our main idea is relatively simple: both generators can undergo degrees of degeneration as their activation levels decrease. Two terms are important in our characterization of the activa-tion levels of the dreamer and the dream environment. The “intensity” of a given conscious or unconscious process is the amount of that process expressed during sleep relative to waking. The “fidelity” of a given proc-ess is the quality with which that process expresses it-self relative to waking. Changes in subjective phe-nomenology across waking and sleep can then be ex-pressed as changes in intensity and/or fidelity of either conscious or unconscious processes, in either the cogni-tive (dreamer) or perceptual (dream environment) do-mains. We apply this thinking first to the dreamer, and then to the dream environment.

THE DREAMER’S COGNITION

Within our framework, the activation of the cogni-tive generator during sleep can range from very low to being equivalent to waking. High cognitive generator activity will result in the waking personality appearing as the dreamer; decreasing cognitive generator activity will result in a dreamer who has less and less fidelity relative to the waking personality, yet who is still com-posed of elements of the waking personality.

We envision the dreamer undergoes something akin to “graceful degradation”, in the neural network sense of Hopfield (1982). “Graceful degradation” occurs in par-allel-distributed networks (PDPs), such as neural nets, as component nodes are removed one by one. Since in-fomation in a PDP is stored through connection weights involving many nodes simultaneously, removal of nodes one by one does not result in abrupt breakdown of net-work fidelity, but a gradual breakdown where informa-tion stored in the network becomes progressively more blurred, imprecise or overlapping (Brown & Zador, 1990). We imagine that the “graceful degradation” of

the dreamer will result from decreased intensity and/or fidelity of unconscious cognitive processors. Thus, looking at changes in conscious cognition during dream-ing should shed light on which unconscious processors are involved in sleep-wake cognitive differences.

In a series of recent papers, J. Allan Hobson, David Kahn, Edward Pace-Schott, and Robert Stickgold have posited a catalogue of changes in consciousness during dreaming. We take this formulation as our point of de-parture for contrasting waking and dreaming conscious-ness. In Table 2, we have developed a composite list of the state-dependent changes in dream consciousness derived mainly from Hobson, Stickgold, and Pace-Schott (1998) and Kahn et al. (1997). Their characteri-zation of the differences in consciousness across states is, in fact, readily interpretable in terms of our notions of perceptual and cognitive generators. In Table 2 we have classified these authors’ differences in terms of four categories. Does the change occur in: (1) the perceptual generator, (2) the cognitive generator, (3) the interaction between perceptual and cognitive generators, or (4) across major brain states (e.g. wake to dream, dream to wake, dream to dream)?

We also need to point out that there is controversy in regards to specific items in Table 2; this list by no means represents a consensus view of dream conscious-ness (Kahan & LaBerge, 1994; Purcell et al., 1993). The view expressed in Table 2 has been termed the “defi-ciency view” of dreaming (Kahan & LaBerge, 1994; Purcell et al., 1993) because it portrays the dreamer as cognitively deficient relative to the waking self. This view further lacks an appreciation of the “multiplicity” of dreams (Hunt, 1989) to which we refer ahead.

In this section we focus on the list of alterations in the dreamer’s cognition presented in Table 2. First, af-fect is generally enhanced, particularly in the negative direction of stress, fear, and anxiety. Correspondingly, “fight or flight” reactions occur relatively more fre-quently in dreams than in waking. Multiple differences in cognition and memory are noted in Table 2.

Generally, these are all deficiencies relative to wak-ing, with the exception of hyperassociativity - an in-crease in associative processing in dream consciousness compared to waking (Kahn et al., 1997). Changes in thinking processes include weak logical rigor, ad hoc reasoning and a substantial incidence of confabulation. Thought is considered delusional in dreams by these authors because “we are consistently duped into believ-ing that we are awake” (Hobson et al., 1998b, Table 1),

18 DeGracia

although it is noted in Hobson et al. (1998b) that lucid dreams are an exception to this feature. Volition is claimed to be weak, and self-reflection (a metacognitive or executive function) is claimed to be absent or greatly diminished relative to waking.

Several changes in the dreamer’s memory are noted: recent memory is diminished, remote memory is en-hanced, and working memory decreased.

A similar deficiency view of dream cognition is cap-tured in Alan Rechtschaffen’s (1978) famous notion of the “single mindedness and isolation” of dreams. Recht-schaffen describes dreams as self-contained experiences, isolated from the on-going episodic experience of the waking self. Within this self-contained experience, the dreamer is completely absorbed in the ongoing sequence of events, lacking the overarching framework of waking

PG-related CG-related PG-CG interactions Across States IO gating: • External perception diminished.

Affect: • Increased emotion, dominated by fear and anxiety.

Attention is lost.

• Internal perception enhanced (e.g. halluci-nations).

• Instinctual programs (especially fight-flight) often incorporated into dreams.

(Conscious PG-CG inter-action).

Memory deficits across dream-wake, wake-dream, and dream-dream transi-tions.

PG defects: • Incongruity (per-ceptual): dream imagery is strange, unusual, or impossible.

Orientation: Thought: • Reasoning ad hoc. • Logical rigor weak. • Processing by hyper-associativity. • Confabulation preva-lent.

• Discontinuity (per-ceptual): discreet or abrupt transition in dream imagery.

• Thought is delusional.

• Semantic uncertainty of the dreamer toward persons, places, and events in the dream. (Conscious PG-CG inter-action).

Volition: • Volitional control weak, greatly attenuated.

Incongruity (semantic): conceptions of persons, places, and things are fused, plastic, incongru-ent, and unstable. (Con-scious PG-CG interac-tion).

Metacognition: • Self-reflection absent or greatly diminished rela-tive to waking.

Discontinuity of dream “plot.” (Unconscious PG-CG interaction).

Memory: • Recent memory di-minished. • Remote memory en-hanced. • Decreased working memory.

Confabulatory narrative construction of the dream “plot.” (Unconscious PG-CG interaction).

Table 2: Comparison of waking and dreaming consciousness taken from Hobson et al. (1998b) and Kahn et al. (1997). We have categorized these authors’ characterizations of across-state differences in consciousness in terms of the no-tions of perceptual generator (PG-related), cognitive generator (CG-related), the interaction of the perceptual and cog-nitive generators (PG-CG interactions) at either conscious or unconscious levels, and across-state differences. The cognitive features in particular represent a deficiency view of dreaming consciousness (Kahan & LaBerge, 1994).

Global Workspace and Dreaming 19

experience to guide the dreamer’s thinking and action. The dreamer appears to be at the mercy of the “relent-lessly unfolding” (Kahn et al., 1997, pg. 17) quality of the dream perceptual environment.

In contrast, some cognitive functions appear unal-tered in dream consciousness. Language during dream-ing, occupying as much as 30% of all dream activity (Hall & Van de Castle, 1966; Meier, 1993), is essen-tially identical to waking. Language in dreams occurs as both overt and covert vocalizations, in the same variety of contexts as during waking. The results of Strauch and Meier (1992) stress the “ordinary, everyday nature” of dream speech. Salzarulo and Cipolli (1974) have shown that the syntactic organization of dream speech is similar to waking speech. The formal characteristics of dream speech include its pragmatic competence, com-plexity, well-formedness, grammaticality, lexical cor-rectness, and syntactic competence (Meirer, 1993). Dreamers also make voluntary choices within the con-text of the dream, reflect on, and pay attention to dream events (Kahan et al., 1997). Dreamers are involved in the execution of goals which are local to the dream cir-cumstances. Although the overall context of waking experience is not present, there is still substantial spon-taneous explicit recall of waking episodic memories by the nonlucid dreamer (unpublished data). Habitual forms of metacognition occur in dreams (Kahan, et al., 1997), which we feel is the basis of being “duped into believing that we are awake” while dreaming; most people do not make a habit of reflecting on their state of consciousness while awake, so why should they when dreaming?

The profile in Table 2 seems to indicate executive cognitive functions are more sensitive than nonexecutive functions to the “graceful degradation” of the cognitive generator (see Benson, 1994, for a comparison of execu-tive and nonexecutive cognitive functions). However, this is a superficial generalization because the average dreamer does exercise executive cognitive functions such as directed attention, volitional actions, metacogni-tive thinking and so forth, but these are all expressed with respect to events occurring in the dream, not with respect to waking experience (Kahan & La-Berge, 1994).

We would suggest that a list of ostensibly formal “cognitive deficiencies” of the dreamer, important as it is for characterizing dream phenomenology, gives a mis-leading impression of the mechanisms of cognitive deg-radation during dreaming. If we look closely at the above profile, what it indicates is that highly automatic

procedural-type actions, such as language use, attention, and even styles of thinking and metacognition retain fidelity relative to waking, but that cognitive activities dependent upon promiscuous access to memory struc-tures stored in the brain show degradation. That is, the content of thinking is faulty, not thinking per se. Thus, we do not feel that executive functions undergo substan-tial degradation during typical normative dreaming, but that mnemonic systems are transformed across the sleep-wake cycle. The claims about memory defects in Table 2, of a diminution of “recent memory” and working memory, and enhancement of remote memory, suggest changes to the processors underlying these mnemonic functions. Interestingly, alterations of memory co-occur with exaggerated emotion; this phenomenology is sup-ported by neuroimaging observations of increased acti-vation of the amygdala (Maquet et al., 1996; Nofzinger et al., 1997) during REM sleep compared to waking. Both the emotional and mnemonic alterations point to changes in medial temporal lobe structures as playing an important role in the functional reorganization of the sleeping brain. Ahead we will elaborate extensively on the role of across-state memory changes when we pre-sent our neurobiological model of dream genesis.

Our view of dream cognition not only addresses de-ficiencies relative to waking, but posits that, at high lev-els of cognitive generator activity, the dreamer will ap-proach the fidelity of the waking personality. Such ex-periences have been termed “lucid dreams” (Hunt, 1989; Green, 1968; Snyder & Gackenbach, 1988). The experi-ences to which we here refer are typically brief, sponta-neous experiences, occurring in the context of night-mares and/or immediately preceding awakening (Green, 1968), suggesting these experiences result from shifts in cerebral activation and neuromodulation as the brain prepares to awaken. As we explicate below, this type of dream lucidity is a “lucid dream” in only the most cir-cumscribed sense. Snyder and Gackenbach, (1988) dis-cuss the difficulties with assessing incidences and fre-quencies, but present the conservative estimates that 58% of the population has experienced this form of dream lucidity at least once in their life, and that, on av-erage, about 13% of an individual’s dreams contain such elements of lucidity.

When the dreamer approaches the fidelity of the waking personality, a consequence is that the dreamer displays memory similar to that of the waking personal-ity. This being the case, dream events become per-ceived with respect to memories of waking experience,

20 DeGracia

and the dreamer can realize he or she is dreaming. In the average circumstance, following this mnemonic al-teration the dreamer awakens. This low frequency case of the waking personality “coalescing”, as it were, within a dream environment, may be considered an “embryonic” lucid dream. We now discuss how this ex-perience provides a phenomenological bridge between lucid and nonlucid dreaming.

LUCID DREAMS

If the comparison of waking and dreaming has been an elusive endeavor, so too has been the attempt to com-pare lucid and nonlucid dreams. Lucid dreams have tra-ditionally been defined as “dreams in which the dreamer knows they are dreaming” (Green, 1968; LaBerge, 1985). After personally spending a considerable amount of time in this state of consciousness, we do not believe this conception adequately reflects the essence of these experiences. Previously, LaBerge presented the notion of a “lucid dream schema” as a means to characterize the cumulative cognitive nature of these experience (La-Berge, 1985). Here we expand this notion and describe what we shall call the “lucid dream context.” In both our opinion and experience, it is the operation of a lucid dream context that necessarily distinguishes lucid from nonlucid dreams. Issues of physiological brain activa-tion remain relevant with regard to the innate variability of lucid dreaming (discussed below). We also discuss the imperative to be sensitive to the degree of develop-ment of the lucid dreamer’s skill base in characteriza-tions of dream lucidity.

A review by Gackenbach (1988) of the content of lucid and nonlucid dreams concludes “lucid dreams are more like nonlucid dreams than different.” The phe-nomenological details of lucid dreams indicate that they partake substantially of psychological elements from both nonlucid dreams and waking (Green & McCreery, 1994; Kahan et al., 1997; Kahan & LaBerge, 1994; La-Berge & DeGracia, 1999). We feel that ideas such as Covello's (1984) characterization of lucid dreams as “an incongruous blend of waking and dreaming”, Mahowald and Schenck’s (1992) idea of the overlapping of dream-ing and waking, or Charles Tart’s (1988) suggestion that the lucid dreamer is the waking self, approach a more accurate characterization of these experiences. The lucid dreamer purportedly has a heightening of those same cognitive functions that are deficient in nonlucid dream-ers but present during waking: voluntary control of ac-

tion, wider voluntary memory access, increased self-reflectiveness, and increased quality of thinking. In con-trast to nonlucid dreams, lucid dreams are generally re-membered by the waking personality following awaken-ing. Stated more precisely, lucid dream experiences con-tribute substantially to the episodic memory structure of the waking personality. Many studies have attempted to correlate overt waking cognitive, perceptual, or person-ality characteristics with the propensity to lucid dream (reviewed in Snyder & Gackenbach, 1988). However, given the imprecise and variable definitions of dream lucidity in these early studies, it has been difficult to draw any firm conclusions on this issue. Perhaps the most robust result is that lucid dreaming has been corre-lated with increased dream recall ability (Gackenbach, 1988; Gackenbach, 1991a, 1991b; Hunt, 1989; Snyder & Gackenbach, 1988; Spadafora & Hunt, 1990).

Such observations, along with REM state neuroi-maging (Braun et al., 1997; Maquet et al., 1996; Nofz-inger et al., 1997) and brain lesion data (Solms, 1995, 1997), have led Hobson et al. (1998b) to suggest that lucid dreaming results from the activation of dorso-lateral prefrontal cortex (DLPFC) and the concomitant appearance, in the dreamer’s cognition, of the executive cognitive functions associated with this brain region (D’esposito & Grossman, 1996; Weinberger, 1993). We have, in part, addressed this, increasing the fidelity of the cognitive generator during sleep will produce ex-periences of “embryonic” lucidity as described above. The major weakness to the suggestion that DLPFC ac-tivity is the sole basis distinguishing lucid and nonlucid dreams is its failure to account for the expression of ex-ecutive functions in nonlucid dreams as discussed above, or the failure of executive functions during lucid dreams (Gillespie, 1983; Gillespie, 1984; Green & McCreery, 1994; LaBerge & DeGracia, 1999). In spite of ostensible differences with nonlucid dreams, many of the same cognitive deficits associated with nonlucid dreams can occur in lucid dreams: lapses in thinking, failure of voluntary memory access, incidences of con-fabulation, decreased rigour of thinking, and difficulty in recalling of lucid dreams after awakening (Hunt, 1989; Levitan 1994; LaBerge & DeGracia, 1999). A lucid dreamer’s cognitive fidelity can vary both within and between lucid dreams. Such variations in the lucid dreamer’s cognition are part of the innate variability of lucid dreaming (LaBerge & DeGracia, 1999) referred to above. Again, these deficits appear to revolve around issues of promiscuous memory access as opposed to ex-

Global Workspace and Dreaming 21

ecutive function. In short, the spectrum of cognitive variation associated with nonlucid dreaming also occurs in lucid dreaming.

THE LUCID DREAM CONTEXT

If lucid dreams are not simply an increase in pre-frontal executive cognitive function (or more generally, enhanced cognitive generator activity) over nonlucid dreams, then what distinguishes lucid and nonlucid dreaming? Hunt (1989) refers to lucid dreaming as the development of the dimension of self-reflectiveness within dreams. We feel a more precise definition along such lines is required:

Lucid dreaming is a line of skill development

initiated and sustained by the waking personality, that seeks to create a framework of belief and ac-tion which serves to bring the waking personality into the dream state on a repeatable basis.

This definition indicates that lucid dreams are less a

specific type of dream, and more a specific type of ex-perience of the waking personality. The mental set that allows this control is what we call the “lucid dream con-text.” We have recently expounded (LaBerge & DeGra-cia, 1999) that the lucid dream context consists of at least three cooperatively acting components operating at metacognitive, declarative, and action/skill levels of cognitive organization (Figure 7).

The reference to state (RTS) is a metacognitive reference to one’s state of consciousness while dream-ing. It is not the mere statement “I am dreaming” made while dreaming, because this statement can be made as simple abstract knowledge lacking the metacognitive component (Hunt, 1989; Tart, 1988). Although the RTS would seem to be an increase in executive functioning (e.g. increased metacognition), the basis for the RTS is the increased fidelity of the lucid dreamer’s voluntarily-accessible memory. It is only by contrast to memories of waking that a dreamer can know they are dreaming (Tart, 1988). Thus, the dreamer’s accessible memory must resemble that of the waking self. The RTS can be explicit, as when one realizes, “I am dreaming.” For experienced lucid dreamers, the RTS is implicit; it forms a contextual basis framing the dreamer’s thoughts and metacognitive reflections (DeGracia & LaBerge, 1999). The comparison of the RTS underlying dream lucidity to the waking experience of transpersonal states of con-

sciousness has been suggested (Gackenbach, Cranson, & Alexander, 1986; Hunt, 1989). We do not agree with this characterization; our more modest experience has been that the waking equivalent of the RTS condition during dreaming is simply the metacognitive realization one is not dreaming while awake.

The declarative framework includes: (1) a set of semantic beliefs used by the waking personality to con-ceptualize lucid dream experiences, and (2) the accumu-lation of episodic memories of these experiences. We have indicated (LaBerge & DeGracia, 1999; LaBerge, Levitan, DeGracia, & Zarcone, 1999) that it is not nec-essary for the waking personality to conceptualize their experiences as dreaming; equally prevalent semantic concepts for this experience include “out-of-body ex-periences” and “astral projections” (DeGracia, 1997). As has been previously argued (LaBerge, 1985; Levitan & LaBerge, 1991), astral projections, out-of-body ex-periences and lucid dreams are all the same phenomena. These terms, we believe, emphasize different aspects of the innate variability of lucid dreams (LaBerge & De-Gracia, 1999). The only substantial distinction between these terms is that they represent different semantic no-tions used by individuals to conceptualize their experi-ence. Semantic beliefs are important because they con-strain what the dreamer believes is and is not possible within the dream state (LaBerge, 1985).

The goal-options framework consists of a set of goal-options, or action/skill choices associated with lu-cid dreaming. There are two categories of action/skills contained in the lucid dream context: (1) the range of behaviors exercised while in the dream state, and (2) the actual practices that allow the waking personality to pull itself into the dream state. The former set of actions is heavily conditioned by the semantic framework (La-Berge & DeGracia, 1999); the latter is relatively inde-pendent of semantic beliefs but not of episodic memory (e.g. compare LaBerge and Rheingold, 1990 to Rogo, 1986).

Examples of action skills voluntarily executed by lucid dreamers include those particular to the dream state - flying, passing through walls, and similar pseudo-magical actions (LaBerge & DeGracia, 1999) - and those carried over from waking, such as reading (albeit with difficulty as described in Green and McCreery, 1994; LaBerge,1985), speaking, walking, etc. Paul Tholey (1988, 1991) has pursued research on the devel-opment of waking skills in lucid dreams. Some actions are unintentional and forced on the dreamer by the dream envi-

22 DeGracia

ronment. Perhaps the most common of these are skills lucid dreamers develop to prevent the loss of lucidity (LaBerge & Rheingold, 1990), and skills to prevent lucid dreams from fading (LaBerge, DeGracia, & Zimbardo, 1999; LaBerge & Rheingold, 1990); the necessity to develop such skills is very significant and is elaborated below.

Other actions carried out by lucid dreamers are con-scious cognitive access and control operations: access-ing waking memories, thinking of dream events in terms of waking experience, making explicit effort to remem-ber lucid dream experiences, and actively paying atten-tion to, and exploring, the dream perceptual environ-ment (LaBerge & Rheingold, 1990).

The methods in the goal-options framework that al-low the waking personality to voluntarily enter the dream state are of particular interest. These methods (LaBerge, 1985; LaBerge & Rheingold, 1990; Rogo, 1983; Price & Cohen, 1988; Ophiel, 1982) include some mix of: (1) cultivating the desire to appear in the dream

state, (2) developing dream recall abilities, (3) a set of concentration techniques that allow the waking person-ality to maintain conscious cognition across the wake-sleep border (as illustrated in Figure 1B), and (4) the practice of techniques that condition one to metacogni-tively reflect on one’s state of consciousness while both waking and dreaming, called “state-testing” (LaBerge and Rheingold, 1990; Purcell, Mullington, Moffitt, Hoffmann, & Pigeau, 1986; Purcell et al., 1993).

Current views generally see dreaming as a passive process; dreams just “happen” to the sleeping self (Empson, 1989). Clearly dreaming is a spontaneous activity of the nervous systems but, like other aspects of our physiology and psychology, dreaming can be con-trolled and refined. All lucidity induction techniques are initiated and sustained by the waking personality. Both the declarative and goal-options frameworks of the lucid dream context are subject to procedural memory effects;

metacognitive reference-to-state

declarative framework

semantic beliefs

episodic memories

of other luciddreams

goal-options framework

induction techniques

PerceptualGenerator

Cognitive Generator

global workspace of consciousness

Lucid Dream Context

other unconscious cognitive processes involved in dream generation

potential attenuation of spontaneous dream process

in-dreamactions

Fig. 7. Structure of the lucid dream context and its relation to the spontaneous process of dreaming. The lucid dream context is a cognitive contextual set developed during waking, and expressed and sustained by the self during dream-ing. Specific metacognitive, declarative (semantic and episodic), and procedural elements are required for the skilled execution of lucid dreaming The lucid dream context results in the voluntary-induced heightening of cognitive activity during dreaming, allowing the waking personality to manifest as the dreamer. The lucid dream context places an in-creased computing load on the cognitive generator during dreams and may potentially withdraw computing resources from unconscious cognitive processes contributing to the structure of the dream perceptual environment. This mecha-nism explains the propensity of lucid dreams to fade.

Global Workspace and Dreaming 23

relevant processors can be fine-tuned and optimized through practice, and lead to the development of skilled and habitual behaviors. Thus, lucidity induction tech-niques, dream recall abilities, in-dream actions, and se-mantic beliefs about the nature of the experiences can all become habitual (form contextual hierarchies) to the experienced lucid dreamer. Lucid dreaming is thus a skill the waking personality can choose to cultivate like any other skill such as playing piano, painting, doing mathematics, etc. (LaBerge, 1980; Moffitt & Hoffmann, 1987).

The lucid dream context is our name for the mental set that allows the waking personality to exercise some degree of voluntary control over the dreaming processes. Stated somewhat differently, in lucid dreaming, the per-sonality seeks to control its own brain activity during sleep and bring itself into the dream. As lucid dreaming techniques require the strong active participation of the waking personality, and undergo the transformation from requiring conscious effort to a relatively uncon-scious routine (next section), this suggests a plastic re-organization of dream neurophysiology via intentional effort. The intentional effort results in a reorganization of mnemonic processes during sleep such that the con-scious cognition of the sleeping brain approaches that of the waking brain. It is this intentional effort that sepa-rates skilled lucid dreaming from spontaneous, “embry-onic” dream lucidity. For developed lucid dreamers, the increased brain activation derives from intentional con-trol superimposed over the dreaming process, whereas embryonic lucid dreams result from the spontaneous and unconscious neurophysiology associated with the sleep-wake transition. The underlying global brain physiol-ogy may turn out to be quite similar in both cases. However, it is the source of brain activation that is sig-nificantly different. Hence, lucid dreaming is defined by us primarily in cognitive terms. This cognitive struc-ture - the lucid dream context - becomes another contex-tual structure in the hierarchy of contexts that make up the waking personality. This is not to say issues of un-conscious brain physiology are unimportant in the phe-nomenology of lucid dreaming, only that they are not the primary distinguishing factor between lucid and nonlucid dreaming.

INTENTIONAL CULTIVATION OF A LUCID DREAM CONTEXT

A lucid dream context does not appear fully formed in the mind of the lucid dreamer. Skilled lucid dreaming

results from a developmental sequence of learning and practice. Like any other skill learning, this development involves a transformation from intense conscious par-ticipation to relatively unconscious expectations and rote skills (Baars, 1988). Because of the imprecision that exists in conceptions of lucid dreaming (Blackmore, 1982), we here explicate the developmental progression of skilled lucid dreaming. We draw on Baars’ (1988) notion of the adaptation cycle as a means to conceptu-alize this developmental process. In Baars’ words:

"In learning about a new source of knowledge we of-

ten start with considerable uncertainty and confusion. By paying attention to the problem, a sense of clarity is often gained, as we become conscious of what is to be learned. Finally, with practice, the material becomes highly pre-dictable and fades from consciousness. These three stages make up what we call the adaptation cycle: Starting only with the knowledge that there is something to be learned, the first stage of context creation is resolved as the ele-ments to be learned are defined; in the second stage we have a working context for understanding new material, which is now informative - that is, input now serves to re-duce uncertainty within the working context. In the third stage, we have adapted completely, and lose conscious access to the learned material" (Baars, 1988, pg. 184).

Thus, according to Baars' account, there are three

phases to learning: (1) context creation, (2) context de-velopment, and (3) adaptation. In the first two phases, consciousness plays a critical role in defining and as-similating the new context, respectively. In the final phase, the context is established and aspects of the con-text become, to a large extent, unconscious mnemonic factors framing conscious operations. We suggest it is essential to view lucid dreaming from this perspective.

The creation phase of a lucid dream context leads to the first experience of the waking personality manifest-ing as the dreamer. An individual can cultivate a sponta-neous RTS experience on a trial-by-error basis, or via instruction, into a lucid dream context. Alternatively, the possibility of lucid dreaming may be introduced at the semantic level by some form of instruction. However, semantic knowledge of lucid dreaming does not lead to formation of a lucid dream context. The important se-mantic information to be acquired at this stage is to learn methods to induce the RTS condition during dreaming. There must be an experiential occurrence of the RTS in a dream at some stage, allowing the waking personality to actually manifest as the dreamer. This experience then

24 DeGracia

serves as a point of nucleation in memory around which a lucid dream context can develop.

Once the waking personality has had a direct ex-perience of being in the dream state, and has learned techniques for intentional RTS induction, there follows an exploratory phase involving practice of the lucid dream induction methodology and the learning of new experiences obtained as a result of these practices; this is the informative, or context development phase. The functional importance of the declarative framework grows during this phase as it provides a basis within which to conceptualize episodic experiences and set ac-tion boundaries within lucid dreams. The goal-option framework also grows as effective and possible behav-iors are identified.

After time, procedural elements of dream lucidity induction become habitual (Worsley, 1991). Repeating features of RTS occurrences in lucid dreams are assimi-lated and begin to transform into unconscious metacog-nitive habits and patterns of expectation. Aspects of lu-cid dream behavior that initially required conscious par-ticipation become unconscious expectations, condi-tioned by the increasingly implicit declarative frame-work. Thus, with the adaptation phase of establishing a lucid dream context, many aspects of the context be-come habitual.

This latter consideration has implications in regard to identifying putative patterns of brain activation asso-ciated with lucid and nonlucid dreams. It has been shown that development of new skills is associated with widespread brain activation, particularly involving pre-frontal cortex, but as the skill becomes more refined, and transfers to unconscious rote, brain activation be-comes more focal, involving only those regions needed to mediate the task (Posner & Tudela, 1997). Such con-siderations may also apply to lucid dreaming. For ex-perienced lucid dreamers who have made the process of lucidity induction habitual, there may be no difference in regional activation patterns between such individuals and nonlucid dreamers undergoing dreams with equiva-lent degrees of global cerebral activation. Confirmation of this possibility would underscore the cognitive, as opposed to biological, conception of lucid dreaming.

One consequence of our view is that lucid dream re-ports need to be analyzed with respect to the develop-mental stage of the lucid dreamer. Reports from experi-enced lucid dreamers will tend to be richer than those of inexperienced lucid dreamers. However, facets of the lucid dream may be implicit in the reports of experi-

enced lucid dreamers, but fail to be recognized in the reports of inexperienced lucid dreamers. Another con-sideration involves sensitivity to subjects’ semantic con-ceptions; lack of such sensitivity accounts for some of the ostensible cognitive deficiencies of lucid dreams (LaBerge & DeGracia, 1999). We believe these consid-erations begin to clarify issues that have confused the study of lucid dreams (Blackmore, 1982) and their rela-tionship to both waking and nonlucid dreaming. In ef-fect, what we are saying is that a lucid dream is not a lucid dream is not a lucid dream. In analyzing lucid dreams, it is critical to be able to pinpoint where along the developmental spectrum a particular subject lies. Different criteria of analysis will necessarily apply at different stages of development. Clearly the cognitive considerations are vastly different between isolated, spontaneous occurrences of lucidity preceding awaken-ing, and the intentionally-induced lucid dreaming we are describing here. The former reflects spontaneous and unconscious neurophysiology, the latter is the result of conscious and intentional development of the skills, conceptions, and experience base outlined above.

Thus, to summarize our view of the dreamer’s cog-nition: this can undergo a spectrum of activation ranging from the low complexity NREM-associated mentation, to the expression of the waking personality and associ-ated forms of lucidity. The aspects of the dreaming self most sensitive to “graceful degradation” appear to be those involved in declarative memory formation and access; the most insensitive aspects appear to be over-learned cognitive activities. As the declarative mne-monic system degrades, this creates the appearance of executive control dysfunction that has been associated with deficiency views of dream cognition. The waking personality can develop a lucid dream context, a set of methods for intentionally taking the waking personality into the dream state relatively intact. This effort is su-perimposed over the innate activity spectrum associated with the cognitive generator such that lucid dreamers can display cognitive deficits similar to nonlucid dream-ers.

THE DREAM PERCEPTUAL ENVIRONMENT

If the dreamer undergoes “graceful degradation” during sleep, the perceptual environment (PE) displays much more abrupt change. Here we outline the phe-nomenology of sleep perceptual environments.

In section 2 (see also Figure 1) we spoke of the spectrum of complexity of the sleep PEs. Given our

Global Workspace and Dreaming 25

present understanding of the cerebral organization of sensory modalities, we can discuss these changes in terms of the presence or absence of specific modalities. Thus, no modalities operate in pure mentational experi-ences. Hypnagogic imagery is characterised by: (1) the presence of visual object modalities minus visual-spatial context, (2) the occasional presence of auditory “object” modalities minus an auditory-spatial context, and (3) the rare occurrence of somatosensory modalities, which more often are not present. In dreams proper, the whole repertoire of sensory modalities exist. Some authors have associated dreaming predominantly with visual imagery, but when we consider that roughly 50% of cerebral cortical area is dedicated to visual processing (Felleman & Van Essen, 1991), then it is no surprise that vision is prevalent in dreaming. The difference between dreamer-as-actor dreams and dreamer-as-observer dreams involves somatosensory spatial immersion in the PE for the former and not the latter.

Content analysis of dreaming provides us with in-formation as to the frequency of occurrence of things in dream PEs. Generally, anything found in the waking PE can exist in dream PEs. The Hall and Van de Castle (1966) content rating scale looks like a categorization of normal waking perception. When we look at the norma-tive content data collected by Hall and Van de Castle, it indicates that dream PEs are much more like waking than the usual idea of dreams as disjointed bizarre con-structions (Hunt, 1989). However, dream PEs contain a broader variety of content than waking. For example, there is a classification for characters that metamor-phize: “It sometimes happens…that a character changes his or her sex, identity, or age…[or]…to change into an animal or vice versa.” This broadening of perceptual content can be considered a form of conscious percep-tual hyperassociativity characteristic of dreaming.

In Table 2 are listed the major changes in the dream PE according to Hobson and colleagues. Stated in our terms, the major change in the perceptual generator across dreaming and waking is that sensory parameteri-zation is diminished and perception of internally gener-ated hallucinations increases. Perceptions of the dream PE are characterized as “bizarre.” Perceptual incongrui-ties are the superimposition of perceptual features from multiple objects onto a single object, and perceptual dis-continuities are the abrupt changes in the dream PE or things encountered therein (Hobson, 1988; Kahn et al., 1997). Kahn et al., (1997) presented the hypothesis that perceptual bizzareness can be thought of as the “defec-

tive” binding (Hardcastle, 1994; Llinás & Ribary, 1993; Von der Malsburg, 1996) of dream perceptions, result-ing from the lack of sensory stabilization over time. The normative data indicates that bizzareness is less frequent than its absence (Hunt et al., 1982), but it is clearly a perceptual phenomena characteristic of dreaming, and not waking, consciousness.

We discussed variations in the lucid dreamer’s cog-nition above. The other important component to the innate variability of lucid dreams is the tremendous variation in lucid dream PEs, of which we have identi-fied three types: “minimal,” “typical,” and “surreal” (LaBerge & DeGracia, 1999). All three types share the feature of being an external perceptual space within which the lucid dreamer is immersed. An example of a minimal PE was described and characterized in section 2. Magallón (1991) has described minimal perceptual environments as “imageless lucid dreams.” Typical dream environments are those of normative nonlucid dream PEs, and resemble waking perception. Surreal PEs are abstract spaces of color and motion that bear no resemblance to waking and nonlucid dream PEs. Hunt et al. (1982) has described surreal environments as psy-chedelic transformations of the formal properties of vi-sion. It should be emphasized that surreal environments are not forms of dream perceptual bizarreness, but are abstract perceptual spaces within which the dreamer is immersed.

There is substantial variation in the expression of minimal and typical PEs across lucid dreams. For exam-ple, any given typical environment may contain more or less perceptual bizarreness, may appear more or less realistic (Green & McCreery, 1994; McCreery, 1973), and can appear hyper-vivid compared to waking percep-tion (LaBerge, 1985). Minimal environments may sometimes contain weak expression of visual form and/or color perception, and the “void-like” space of these dreamscapes can take on a variety of qualities or textures (LaBerge & DeGracia, 1999). To speak of variability of surreal environments makes no sense be-cause what is perceived in these environments is not even comprehendible by the lucid dreamer (LaBerge & DeGracia, 1999).

The differences between mentational, hypnagogic, and nonlucid dream PEs have served to distinguish these experiences, along with the statistical association of each with specific sleep stages. The parallel between these and the three lucid dream PEs is intriguing. Minimal environments are reminiscent of the lack of a

26 DeGracia

PE during mentational experiences. Visual imagery is abstract by definition in surreal environments, but hyp-nagogic imagery tends in this direction as well. What distinguishes these pairs is a definite sense of immersion in the lucid dream-associated PEs. Typical dream envi-ronments, which are very much like waking perceptions, occur in both lucid and nonlucid dreams.

What is truly interesting is that different stable states of perceptual consciousness are at all possible. From our perspective, these variations reflect different states of activity of the perceptual generator. Based on the idea that chaotic dynamics can model large scale neuronal assembly behavior (Freeman, 1994; Freeman & Skarda, 1985; Globus, 1992, Hardcastle, 1994), a reasonable speculation is that the different PEs could correspond to discreet attractor states of cortical-subcortical perceptual activity, and represent major stable phases of conscious perceptual output. Our personal experience with surreal environments suggests they may be semi-stable attractor states of perception falling between minimal and typical environments (LaBerge & DeGracia, 1999). The ab-stract and geometric imagery of surreal and hypnagogic environments, reminiscent also of psychedelic drug-induced hallucinations (Aaronson & Osmond, 1970), may be the direct conscious perception of neurocompu-tational processing stages that are normally unconscious, but underlie the content of typical environments and normal waking perception (Hunt et al., 1982; Mavroma-tis, 1987).

INTERACTION OF THE DREAMER AND DREAM ENVIRONMENT IN NONLUCID DREAMS

Above we discussed that the interactions between dreamer and PE vary across types of sleep experience ranging from none (mentational experiences), to rela-tively passive observation (hypnagogia and dreamer-as-observer dreams) to full perceptual-motoric immersion (nonlucid and lucid dreaming). Clearly an important basis for these types of dreamer-PE interactions is the repertoire of modalities present in the dream. For ex-ample, if the dreamer does not possess somatic modali-ties or somatic-spatial immersion within the PE, then interactions will necessarily exclude somatic-motor ac-tivity, and the dreamer can only “passively” watch the dream perceptions.

In immersion dreams, several of the features of dream consciousness identified by Hobson and col-leagues can be construed as representing interactions between the dreamer and the dream PE (Table 2). As

illustrated in Figure 6, our framework indicates that the interaction between the dreamer and the dream PE can occur via either conscious or unconscious information flow pathways, and that the direction of the interaction can be from the cognitive to the perceptual generator or vice versa. We now apply this logic to the dreamer-PE interactions listed in Table 2.

Hobson and colleagues claim that attention is lost during dreaming. One wonders how the dreamer’s thoughts with respect to the dream PE could be so well organized if the dreamer was not actively paying atten-tion to the dream PE. In fact, a dream quoted in Kahn et al., (1997) shows the dreamer consciously attending to several details of the dream environment. We have dis-cussed above that dreamers indeed pay attention to the dream environment, and this can be construed as a cog-nitive-to-perceptual interaction occurring in conscious-ness.

Hobson and colleagues suggest instability of the dreamer’s mental orientation is a formal feature of dream consciousness. This is true in the sense that the dreamer will tend to display more mental orientational instability than the waking self, but all dream cognition is not characterized by this property. Orientational in-stability encompasses two distinct processes: (1) a se-mantic uncertainty whereby the dreamer is uncertain of the identity of persons, places, or things in the dream, or (2) semantic incongruities whereby conceptions of persons, places, and things fuse and/or are unstable. Such factors represent the dreamer’s semantic interpre-tation of the dream PE, and can thus be construed as consciousness cognitive-to-perceptual interactions. Kahn et al. (1997) offers the hypothesis, similar to that of Antrobus (1986), that instability of mental orientation may be due to decreased working memory capacity. Decreased working memory would prevent the dreamer from holding perceptions of the dream environment in working memory long enough to detect errors between perceptions and semantic responses, thereby leading to instability of mental orientation. We note that such a hypothesis implies alterations in the unconscious func-tioning of working memory. We also note that this fea-ture of dream consciousness is again related to the promiscuity of consciously accessible memory.

Discontinuity of the dream “plot” is noted; disconti-nuity at this level of the dream’s organization would be expected to directly follow dream perceptual disconti-nuities (Montangero, 1991), suggesting to us a percep-tual-to-cognitive generator interaction. Dream disconti-

Global Workspace and Dreaming 27

nuities are in no obvious way related to the dreamer’s conscious cognition or intentions, are grounded in brain neurophysiology (Hobson & McCarley, 1977; Kahn & Hobson, 1993), and hence can be construed as uncon-scious events. In nonlucid dreams, perceptual disconti-nuities are not recognized as such by the dreamer, but they alter the dreamer’s conscious cognition such as goal orientations or mental set (again see the dream ex-ample provided in Kahn et al., 1997), suggesting that perceptual discontinuities serve as a trigger to initiate transformations of the contextual hierarchy framing the dreamer’s cognition.

Finally, it is stated that there is confabulatory narra-tive construction of the dream “plot.” This is a very ambiguous statement, attributing unintentional actions to the dreamer, and implying a formal narrative construc-tion of dreams. Confabulation is the unintentional generation of falsehoods, is common with brain damage that upsets prefrontal circuits (Johnson, O’Conner, & Cantor, 1997), and is hypothesized to be due to different types of memory dysfunction such as faulty retrieval or faulty error checking (Schacter, Norman, & Koutstaal, 1998). Our evaluation of nonlucid dream reports sug-gests that some forms of dream confabulation are, in fact, incongruous associations of waking memories (un-published data), implicating faulty retrieval mecha-nisms. The notion that the dreamer unintentionally gen-erates the dream plot is equivalent to saying the dreamer unintentionally generates the dream. Thus, this sup-posed “feature” of dream consciousness actually implies a mechanisms of dream generation. We return to this issue in section 8.4, when we present a model of dream generation which offers a definite relationship between the dreamer and the dream environment at both con-scious and unconscious levels of cerebral function.

Again, if we look closely at the above interactions, we see patterns emerge consistent with changes in the dreamer’s cognition and perception discussed above. Perceptual-to-cognitive interactions tend to be abrupt and propagate through the dream by altering uncon-scious contextual levels of the dreamer’s cognition. The typical nonlucid dreamer spontaneously responds to per-ceptual discontinuities by altering mental orientation, goal objectives and other elements indicative of the op-eration of cognitive context hierarchies. The cognitive-to-perceptual interactions tend to reflect dysfunctions in promiscuously accessible memory systems resulting in mental disorientation or confabulation by the dreamer.

INTERACTION OF THE DREAMER AND DREAM ENVIRONMENT IN LUCID DREAMS

Dreamer-environment interactions become particu-larly discreet during lucid dreams. Lucid dreamers come to learn that their conscious thoughts are not, in any obvious fashion, creating the dream perceptual envi-ronment. The lucid dreamer’s mental orientation is not one of actively creating the dream environment, but is more akin to active exploration of the PE (LaBerge & Rheingold, 1990). Lucid dreamer’s interact extensively with the PE, and many discussions of lucid dreaming have characterized “dream control” as an important ele-ment of lucid dreaming (Moffitt, Hoffmann, Mullington, Purcell, Pigeau, & Wells, 1988; Purcell, et al., 1993; Schwartz & Godwyn, 1988; Tart, 1988).

However, there has been less discussion and focus on the limits of the lucid dreamer to control the dream environment. These limits are substantial, and provide important clues to basic processes underlying dream generation. The content (places and objects) of the PE is generally not arbitrarily manipulatable by the lucid dreamer in the manner imaginative imagery is (e.g. see example in Kahan and LaBerge, 1994; Worsley, 1988). Nor does the lucid dreamer intentionally create the dream PE; it spontaneously “appears” as it does in nonlucid dreams (Green & McCreery, 1994; Hartmann, 1973). Further, events in the PE occur independent of, and often contrary to, the conscious intentions and ex-pectations of the lucid dreamer. Examples include: (1) the “light switch phenomena” (Hearne, 1981; Green, 1968), (2) experiences of uncontrollable kinesthetic sen-sations that manifest as bizarre motor experiences, such as being uncontrollably whisked along by the “wind” (LaBerge & DeGracia, 1999), (3) the inability to per-form magical actions such as flying or passing through walls, in spite of having done so in previous lucid dreams (LaBerge & DeGracia, 1999), (4) being chased by unintended hostile dream characters (Tholey, 1988), or (5) finding the dream environment transform in un-predictable and unintended ways, e.g. perceptual discon-tinuities (Gillespie, 1984).

Above we spoke of the innate variability of lucid dreaming in terms of variation in the dreamer’s con-scious cognition and variations in lucid dream percep-tual environments. The interactions between the lucid dreamer and the dream PE are highlighted in the meth-ods lucid dreamers use to cope with this innate variabil-ity. These methods include: (1) some degree of control

28 DeGracia

over which type of PE they occupy, (2) the ability to “stabilize” their presence and prevent fading of the PE, and (3) the ability to prevent the loss of their lucidity.

Lucid dreams possess a propensity to “fade” (La-Berge, DeGracia, & Zimbardo, 1999). Fading is related to an important, and perhaps underappreciated, relation-ship between the lucid dreamer and the lucid dream PE: lucid dreamers learn that they can consciously sense how stable or unstable their presence is in the dream PE (LaBerge & DeGracia, 1999; Gillespie, 1984; Hunt, 1989). That is, the lucid dreamer may “feel” very stable within the dream perceptual environment, or may “feel” very unstable, with a continuum of variation between these extremes. When a lucid dream is “unstable”, the lucid dreamer’s vision may phase in and out of blind-ness, colors may become pale and washed out, somes-thetic sensations may weaken (for example: Lischka, 1979; Marcot, 1987; both quoted in Green & McCreery, 1994). If dream perception becomes too unstable, the dream PE simply fades, and the lucid dreamer usually awakens in full consciousness. The loss of perceptual modalities during lucid dream fading is opposite that described in section 2 for the onset of the perceptual environment, with vision being the most sensitive to disruption (LaBerge, DeGracia, & Zimbardo, 1999). Sensations of stability occur in all three types of lucid dream PEs. It is significant to note that such sensations do not occur in the nonlucid dreamer’s consciousness, with the occasional exception of the sensation of fading from the dream concomitant with awakening. In con-trast, sensations of perceptual instability can occur through an entire lucid dream, and vary substantially amongst lucid dreams. The most effective methods for stabilizing the lucid dream PE and preventing it from fading involve enhancement of conscious somatic sensa-tions (LaBerge, DeGracia, & Zimbardo, 1999).

It is also recognized that lucid dreams possess a pro-pensity to revert to nonlucid dreams (Hunt, 1989; Green and McCreery, 1994; LaBerge & Rheingold, 1990). A close look at such transitions reveals that the lucid dreamer becomes absorbed in a spontaneously unfolding dream “narrative” which distracts and absorbs the lucid dreamer’s attention and thought processes (Green & McCreery, 1994; Hunt, 1989, pg. 120; DeGracia & La-Berge, 1999). This transformation, we suggest, can be conceptualized by Baars’ notion of competition amongst contextual hierarchies. Competition occurs between the spontaneous dream process and the lucid dream context for control of the conscious cognition of the dreamer.

This competition process manifests as the spontaneous dream process attempting to absorb the dreamer’s atten-tion, intentions, and thought processes by drawing the dreamer into a series of spontaneously unfolding dream perceptual events (DeGracia & LaBerge, 1999; Gilles-pie, 1984). The lucid dreamer must exert intentional effort to overcome being absorbed by the unfolding dream sequence and maintain the lucid dream context. A successful lucid dream is a balance of participating in the unfolding dream events, yet maintaining vigilance to not lose lucidity (LaBerge, 1985; Price & Cohen, 1988). If, however, events in the dream environment over-whelm the dreamer’s attention and thoughts, then these events come to frame the conscious cognition of the dreamer, displace the lucid dream context, and the lucid dream reverts to a nonlucid dream. In fact, this competi-tion mechanism explains the observation that a lucid dreamer can phase in and out of lucidity during a single lucid dream (LaBerge & DeGracia, 1999).

The above perceptual and cognitive limits clearly indicate that lucid dreaming is a nontrivial balancing act. There is a fine line between allowing the brain to be-come too activated (and hence awake), or relaxing vigi-lance so much that the spontaneous dream process dis-places the lucid dream context from structuring the lucid dreamer’s conscious cognition. Lucid dreamers are thus perhaps more at the mercy of the “relentlessly unfold-ing” nature of the dream than are nonlucid dreamers (Green & McCreery, 1994).

The relationship between cognitive lucidity, percep-tual environment type, and the stability of the perceptual environment is neither simple nor clear. One can be highly lucid, but feel either very stable, very unstable or any degree in between. However, when cognitive lucid-ity is lost, and the lucid dream reverts to a nonlucid dream, the perceptual environment generally becomes “typical”, and the nonlucid dreamer has no awareness of “stability,” suggesting that there must be some relation-ship between cognitive lucidity and the dream percep-tual environment. Our model of the generation of the dream environment will offer an explanation for these phenomena (section 8.6).

DREAM GENERATION

We have outlined the key cognitive and perceptual subjective phenomenology of sleep experiences. Our analysis has suggested that key psychological differ-ences across waking and sleep revolve around issues of memory. We now combine these observations and hy-

Global Workspace and Dreaming 29

pothesize a model of dream production in the sleeping brain. We build this model in stages. First we address the phenomenology of dream types to ascertain the memory sources of dreaming. We then present a neuro-biological model of the generation of the dream PE in terms of current findings in the neurobiology of sleep, memory, and perception. Next is presented a model of dreaming in terms of information flows in the perceptual and cognitive generators. After building our model we comment on its implications with respect to dream func-tion in section 9.

MEMORY SOURCES AND DREAMS

Theories of dream production and memory are inex-tricably interwoven. Freud (1900) saw dreams as dis-guised wish fulfillment, implying that memories of goals and desires structured dreams. Recent authors’ concep-tions have been influenced by modern concepts of mem-ory. Bosinelli (1995) has invoked “long-term memory” as a source of dream patterning. Studies by Cavallero and colleagues have categorized memory sources as autobiographical episodes, abstract self-references, or semantic knowledge (Cavallero, 1987; Cavallero, Foul-kes, Hollifield, & Terry, 1990; Cavallero, Cicogna, Na-tale, Occhionero, & Zito, 1992; Cicogna, Cavallero, & Bosinelli, 1991). Antrobus (1986) has drawn upon the memory structure of ACT*, with its working, long-term, and behavioral forms of memory, to model dream pro-duction.

A comprehensive view of the role of memory in dreaming has remained elusive, not the least because of the variegated roles memory plays. Sleep itself is asso-ciated with a profound anterograde amnesia (Roth, Roehrs, Zwyghuizen-Doorenbos, Stepanski, & Wittig, 1988), and the dreams we do spontaneously recall pro-vide only a tantalizing glimpse into our nighttime mental experiences. It seems intuitively obvious that dreams are in some sense drawn from the memory structure of the waking personality. Yet the substantial novelty in dreams indicates they cannot simply be exact reproduc-tions of waking memory. Neither does the dreamer “imagine” or intentionally create the dream environment (Green & McCreery, 1994). As discussed above, the conscious mnemonic operations of the dreamer are al-tered compared to the waking self, and some apparent cognitive deficiencies can be reduced to alterations in promiscuously-accessible memory fidelity across states. Our distinction between lucid and nonlucid dreams re-lies exclusively on differences in memory content and

activity, and how this affects the entire personality/CNS. Studies attempting to correlate sleep stage changes with the quality of waking learning and memory performance have not contributed substantial illumination on the role of memory in dreaming (Purcell et al., 1993). All of this is further complicated by current models, many derived from animal studies, linking the function of sleep to a variety of forms of cerebral information processing (Stickgold, 1999).

Clearly, memory is not a unitary phenomena. Dif-ferent forms of memory can be classified as either de-clarative or procedural and, as we argued above, contex-tual; each of these categories has further subdivision (Gabrieli, 1998; Gaffan, 1994). Moreover, specific cog-nitive operations depend on different memory systems to varying extents. We can ask questions about “within-dream” utilization of memory: what memory sources are used in construction of the dream environment? What sources of memory are utilized by the dreamer during cognitive activities in the dream? We can inquire into “across-state” memory: how is the memory utilized in dreams related to waking memory? How do dreams affect or alter waking memory? We can ask these ques-tions for declarative, procedural, and contextual forms of memory. It is therefore imperative to be as precise as possible when discussing the putative roles of memory in dreaming. When we speak here of sources of memory “structuring” the dream, there are two distinct levels to which we refer: the structuring of the dream PE, and the contextual structuring of the dreamer’s conscious cogni-tion. We focus now on the former.

The structuring of the dream environment is a spon-taneous activity of the CNS, and occurs outside of the conscious operations of either the waking self or the dreamer. That is to say, the generation of the dream environment is grounded in unconscious process-ing. Within our framework, the dream environment is the output of the perceptual generator. The central as-sumption to our conception of dream genesis is that, for the perceptual generator to produce conscious per-ceptual experiences, it must have a source of pa-rameterizing input. Sensory input serves this function during waking. But what serves the equivalent function during sleep?

We suggest that, during dreams, patterns of relation-ship derived from the declarative memory system serve to parameterize, or structure, the conscious output of the perceptual generator. Or more precisely, a flow of in-

30 DeGracia

formation from the MTMS serves as surrogate sensory input to perceptual thalamo-cortical circuits to generate the dream PE. This transaction of information repre-sents the first stage in dream genesis. Subsequent stages will also be outlined, allowing us to clarify the relative roles of the dreamer and the dream environment in the dream production process. We now develop these ideas.

Harry Hunt (1989) has interpreted the phenomenol-ogy of dreaming to indicate that there are multiple types of dreams, and that this typology follows a psychologi-cal, or content-specific type distinction. We feel that that Hunt’s characterization of the multiplicity of dreams is perhaps the most comprehensive formulation of dreaming available, and thus can provide us with the broadest view of memory sources of dreaming. The following is a list of nine dream types identified by Hunt, which we interpret as representing different sources of memory serving as focal points from which different types of dreams emerge:

1. Dreams of verbal metonymy. Verbal meton-ymy is the causal semantic relationship between words, suggesting that this type of dream is related to the causal linkage of experiences within episodic memory, and the description of experience in linear verbal terms. Dreams of verbal metonymy are the normative dreams observed in the sleep laboratory and in home dream reports, and are overwhelmingly life-like. Such dreams derive from the episodic memory system (Tulving, 1972) that re-cords everyday experience.

2. Dreams of verbal metaphor. This dream type refers to dreams in which content is based on linguistic processes not obviously related to episodic experience, but reflects instead the innate structure of the linguistic processors themselves, such as semantic equivalence, clang association, phonetic similarity, rhyme and pun. Such dreams literally seem to emerge from a play on words (Hunt, 1989). We suggest that linguistic proces-sors, dissociated from everyday episodic or pragmatic usage, are acting as the source of memory for structuring dreams of verbal metaphor.

3. Objective dreams. Hunt (1989), following Jung (1961), recognizes some dreams as “outwardly di-rected,” “as surprisingly accurate and/or creative depic-tions of external matters” and associates these types of dreams with “creative breakthroughs in the sciences and the arts.” We would suggest that such dreams are also structured by the episodic mnemonic system, yet biased heavily towards goal contexts and motivational factors.

4. Culture-pattern dreams. This type of dream derives particularly from anthropological investigations of dreaming in native cultures (Cawte, 1984; Tedlock, 1987). It is the dream type of the Shaman who uses dreams as a divinatory or healing tool to fulfill his or her cultural role. We feel that such dreams are based on the semantic memory system (Tulving, 1972), with a heavy emphasis on cultural roles, mores, myths, and customs.

5. Archetypical dreams. Archetypical dreams are associated particularly with the psychology of Jung (Kluger, 1975). These dreams are characterized by “their uncanny-numinous quality and aesthetically rich structure...the powerful sense of felt meaning and por-tent” (Hunt, 1989, pg 129). These dreams seem to be structured by “intuitive” memories, a coupling of emo-tional feelings of profundity and relatively unexplicated insight.

6. Nightmares. Hunt defines nightmares less as a specific dream type and more as a degree of dream in-tensity. Clearly nightmares are characterized by inten-sity of negative affect and can lead to forms of dream lucidity. It appears to us, however, that nightmares are drawn from sources of memory closely associated with fear, fright, and other forms of negative affect

7. Visual-spatial dreams. Hunt suggests that, in some circumstances, the visual-spatial aspects of dreams are their own expression independent of episodic or lin-guistic factors. This line of thought may be particularly true of hypnagogic imagery, and what we have termed “surreal” PEs. These dreams likely represent direct in-trusions of lower level, normally unconscious, forms of visual processing into consciousness. Which is to say, the source of memories from which these dreams are drawn is the innate structuring of the visual system.

8. Titanic dreams. Hunt coined the term “titanic dream” based on Silberer’s (1917) recognition that some dreams contain very strong somatosensory elements. These include strong kinesthetic sensations, forms of mutilation, dismemberment, physical attacks, intense sexual activity, sensations of suffocation, paralysis, fly-ing, floating, falling, and spinning, as well as expres-sions of natural forces such as wind, tornadoes, fire, and water. Analogously to visual-spatial dreams, we suggest that the memory source underlying titanic dreams is the innate structuring of the somatosensory-motor system.

Global Workspace and Dreaming 31

9. Medical dreams. Medical dreams are, according to Hunt, “those in which the dream directly presents so-matic and medical conditions of which the individual may not be consciously aware” (Hunt, 1989, pg. 111). Several studies suggest that specific dream content ac-companies specific medical conditions with more or less reliability (Heather-Greener, Comstock, & Joyce, 1996; Levitan, 1976; Smith, 1986). Hunt attributes such dreams to “background tactile-kinesthetic patterns.” We would suggest that such dreams represent the structuring

of the dream by somatic and/or viserotopic (Loewy, 1990) representations.

What Hunt’s view of dreaming implies is that dif-ferent loci of memory serve as sources of surrogate sen-sory input and provide “parameters” that structure the dream PE. The nine dream types above can be divided into two broad categories (Figure 8): (1) those in which modules we associate with the cognitive generator are parameterizing the perceptual generator (CG-structured dreams), and (2) those in which modules within the per-ceptual generator provide parameterization to the per-

CG-Structured Dream Types

GW

CG PG

YX

CG output

PG output

PG-Structured Dream Types

GW

CG PG

YX

CG output

PG output

Dream Types Dominant Mnemonic Loci Tentative Localization CG-Structured Dreams Declarative subtypes: X = 1. Verbal metonymy Episodic memory MTMS 2. Objective Goal representations MTMS 3. Culture pattern Social representations/roles MTMS 4. Archetypical Intuition/semantic memory MTMS 5. Nightmares Affect MTLM /Amygdala 6. Verbal metaphor Linguistics processors Wernicke’s area PG-Structured Dreams Y = 7. Visual-spatial Visual processors Visual cortex 8. Titanic Somatosensory processors Somatosensory cortex 9. Medical Somato-viscerotopic representa-

tions Somatosensory cortex, brainstem viserotopic maps?

Fig. 8. Different mnemonic loci in either the cognitive (CG-structured) or perceptual (PG-structured) generators substi-tute for sensory input parameterization to the perceptual generator during dreaming. The accompanying table shows the dominant loci associated with each of Hunt’s (1989) dream types. A tentative neuroanatomical localization associated with each category is provided. GW = global workspace, MTMS = medial temporal lobe memory system.

32 DeGracia

ceptual generator (PG-structured dreams). We have classified four of the CG-structured dreams as subtypes of declarative memory because their memory sources are clearly encompassed within this distinction. A core notion to Hunt’s taxonomy is that some dream types reflect specific information content within the nervous system, while others reflect the structure of neurocogni-tive processing relatively independent from specific con-tent. We agree with this general notion; the dream type “verbal metaphor” and all of the PG-structured dreams represent the latter category.

Our classification of Hunt’s typology is illustrated in Figure 8 where the diagrams and accompanying table illustrate the major sources of memory within the cogni-tive (X) or perceptual (Y) generators providing the dominant structuring input to the perceptual generator. We explicitly posit that the input information structuring the perceptual generator flows along unconscious proc-essing pathways; this is depicted by the arrows entering and exiting the bottom of the generators. The outputs from the tops of the cognitive and perceptual generators are depicted as forming the global workspace of the dream. A neuroanatomical assignment of each memory loci is provided and will be referenced ahead.

In presenting these classifications, we also agree with Hunt on the following critical point: “…all dimen-sions of dream formation may be more or less nascent in all dreams…the dreaming process would selectively ex-aggerate one or more of these dimensions…” (Hunt, 1989, pg. 101). Thus, our claim is that specific loci of memory have an enhanced influence on the perceptual generator, leading to the generation of the different dream types. The types are not mutually exclusive. This is indicated in Figure 8 by the dashed lines, showing that both PG- and CG-structured dreams have the same for-mal structure. In our terms, this means that all uncon-scious processes necessary for the action of perceptual and cognitive generation operate in all dream types, but that specific loci of these processors divert activity from conscious cognitive or perceptual output to serve an un-conscious role in parameterizing perceptual generator output.

Since the dominant unconscious cognitive or per-ceptual loci ultimately serve to structure conscious dream perception, we can say that normally unconscious forms of waking information take conscious form as the dream environment (Malamud, 1988). This is not the mysterious subconscious mind envisioned by Freud, because we are referring to precise forms of memory in

the CNS. In this sense then, during dreams, the global workspace - consciousness - becomes a type of “theatre” within which unconscious information manifests in the perceptual consciousness of the dreamer.

THE GENERATION OF THE DREAM ENVIRON-MENT

A number of significant question are raised by the above analysis. How precisely does mnemonic informa-tion substitute for sensory input parameterization and lead to the genesis of the dream PE? How does the overt conscious activity of the dreamer relate to the uncon-scious activities responsible for parameterizing the per-ceptual generator? What is the relationship between the dreamer and the PE? Who creates whom? Where does the “intention” or impetus behind dreaming come from?

To accurately answer these questions, we would need to know the precise time course and patterns of regional brain activation leading to dream generation, and this does not yet exist. The temporal view of sleep provided by EEG analysis does not allow assignment of regional cerebral activation patterns (Hobson et al., 1998a). The activation-synthesis model was expanded to include changes in hippocampal function during REM (Hobson & Schmajuk, 1988), but does not provide a specific pathway of forebrain activation. Neuroimaging studies of the REM brain provide a static snapshot of cerebral activation patterns but cannot yet approach the temporal resolution required to visualize dream onset. Neuropsychological data also offers a snapshot view of which brain regions mediate and support dreaming. Mark Solms (1997) has extrapolated this data to present what he terms a “tentative model” of brain activation underlying dream generation.

The model put forward by Solms invokes a “regres-sive activation” from frontal to parietal lobes as central to the dream generation processes. By this he is refer-ring to a reverse flow of information in the cerebral cortex. Solms conceptualizes the forward waking flow of cortical information to be from posterior sensory cor-tices forward to frontal cortex and hence motor output. Admittedly, this is an oversimplification because the idea of a linear flow from sensory input to motor output has been superseded by a view of complex parallel proc-essing architecture in the brain (Felleman & Van Essen, 1991). Nonetheless, at a coarse-grained resolution, Solms’ notion of “regressive activation” is relevant. It invokes the bidirectionality of intracortical circuits to posit that reverse information flows dominate the dream-

Global Workspace and Dreaming 33

ing process. Echoing a Freudian-type view of dreaming, Solms postulates that the “initiator” region of dreams involves frontal structures related to “appetitive inter-ests.” To prevent awakening via motor command output (the normal waking information flow), frontal activity is “deflected” in the reverse direction and activates multi-modal inferior parietal lobe. The spatial-kinetic and “symbolic-integrative” aspects of dreaming derive from activation of right and left inferior parietal lobes, respec-tively. This frontal-to-parietal activation is structured by frontal “appetitive interests,” providing a Freudian-like “wish fulfillment” dream. Solms considers the rele-vant frontal and parietal regions “essential,” meaning that lesions in these areas abate dreaming altogether.

Okuma (1992) has also posited an activation path-way for dreaming. In this model a stream of vague and disorganized thinking begins during NREM and, with the onset of REM sleep, draws out sensory images from memory to generate the dream PE. Unlike Solms, and following the activation-synthesis model, Okuma re-views data indicating that the activation source for dreams is random and nonspecific activation of the cere-bral cortex and hippocampal system via brainstem neu-rotransmission. From this nonspecific activating input, perceptual imagery is drawn out of memory by a process of free association with the thought-like activity. Stat-ing Okuma’s model in our terms, the conscious output of the cognitive generator (covert or inner speech) struc-tures the perceptual generator (via free association) and neither activates the other. In contrast, the model we now present explicitly indicates that unconscious activ-ity in the cognitive generator structures the perceptual environment and supports conscious sleep cognition.

A NEUROBIOLOGICAL MODEL OF DREAM GENE-SIS

Our model is predicated on: (1) a fundamental re-versal of information flow in the sleeping compared to the waking brain, but in different cortical circuits than those hypothesized by Solms, and (2) on Okuma’s no-tion, derived from the activation-synthesis model (Hobson & McCarley, 1977) of nonspecific activation of hippocampal and cortical circuits. We suggest that the dream perceptual environment is parameterized by the reverse flow of information from the MTMS to the mul-timodal sensory cortex afferent to this system. That is, the circuits that, during waking, mediate the storage of long-term declarative memory - the MTMS - operate in

reverse and activate multimodal cortical representations of the dream PE.

The sequence of events by which we envision the dream PE to be generated are depicted in Figure 9. Changes in brainstem neurotransmission during sleep result simultaneously in: (1) nonspecific activation of the rhinal cortices, via the ventral cholinergic/basal forebrain pathways (Kahn et al., 1997; Kitt, Mitchell, DeLong, Wainer, & Price, 1987) (2) nonspecific activa-tion of higher order thalamo-cortical perceptual systems, via the dorsal cholinergic pathways (Hoover & Baisden, 1980; Jones & Cuello, 1989; Kahn et al., 1997; Rye, Saper, Lee, & Wainer, 1987), and (3) attenuation of the forward flow through the MTMS via enhanced cho-linergic (Ridley, Harder, & Baker, 1996) and decreased aminergic neurotransmission (Harley, 1987, 1991, 1998; Quartermain, 1983). Studies of the REM state brain have consistently shown activation of the parahippo-campal and/or entorhinal cortices (Braun et al., 1997; Maquet et al., 1996; Nofzinger et al., 1997; Ravagnati, Halgren, Babb, & Crandall, 1979), and areas afferent, particularly anterior cingulate gyrus, but also inferior temporal gyrus (Braun et al., 1997), medial prefrontal cortex, and insular cortex (Braun et al., 1997; Nofzinger et al., 1997). Because the rhinal cortices are nodes within bidirectional circuits, we envision that both for-ward and reverse conduction will propagate from ran-domly activated sites in the rhinal cortices.

The forward propagation would enter the hippo-campal system via the dentate gyrus, and such activity may be the basis for hippocampal theta rhythms ob-served during REM sleep and associated with the orient-ing response in awake animals (Buzsáki, 1989; Morri-son, 1983). During waking, the orienting response is associated with forward information flow through the hippocampus, and with enhanced learning (Buzsáki, 1989; Chrobak & Buzsáki, 1998). However, sleep is a state of almost total amnesia (Roth et al., 1988), suggest-ing an attenuation of forward flow into the hippocampal system. The forward flow through the hippocampus would be expected to be attenuated by aminergic de-modulation, which is at a maximum in the REM brain (McCormick, 1992), thereby attenuating a system im-portant for long-term memory formation in the waking brain (Berridge, Page, Valentino, & Foote, 1993; Has-selmeo, 1995; Izquierdo & Medina, 1995; Izumi & Zorumski, 1999; Quartermain, 1983).

34 DeGracia

Reverse propagation from the rhinal cortices would be expected to spread from entorhinal to parahippocam-pal and perirhinal cortices and thence to their afferent sites in widely separated multi- and supramodal cortical regions. This backpropagation of information from the MTMS, we suggest, serves as a source of “surrogate sensory input” and parameterizes, structures, or gives content to, multimodal and higher order unimodal sen-sory cortices to generate the dream PE. We have stated that this pathway is unconscious, which means it is not in any obvious way manipulatable by conscious access and control operations. We now more precisely define the nature of this information.

Autoassociative neural network models of hippo-campal function offer clues as to the nature of this in-

formation. Modeled as an autoassociative network, the hippocampal system must be capable of accepting any arbitrary pattern of inputs (Gluck & Myers, 1997). Spe-cifically, Myers, Gluck, and Granger (1995) suggested that entorhinal cortex may perform a redundancy com-pression between different modalities, or across poly-modal features of a stimulus, similar to Eichenbaum and Bunsey’s (1995) suggestion that entorhinal cortex “fuses” coincident stimuli. In the formation of long-term explicit memories during waking, this “com-pressed” input then enters dentate gyrus where it is sub-ject to the forward hippocampal pathway through re-gions CA3 (the autoassociator, see Gluck & Myers, 1997) and CA1, and ultimately back to afferent sites in the cerebral cortex (section 4.4) where it is encoded as

Entorhinal cortex

Dentate gyrus

CA3

CA1

Perirhinal cortex

Parahippocampal cortex

Cortex afferentto rhinal cortices

releventthalamic nuclei

X

nonspecific, random activating input

(ventral cholinergic system?)

source of hippocampaltheta?

}

} backpropogationof entorhinal cortexinformation

}attenuation of forward hippocampal flow(aminergic demodulation?)

}paramaterizationof multimodal thalamocortical loops

concurrentthalamocorticalactivation

}nonspecific, random

activating input(dorsal cholinergic

system?)

Fig. 9. A neurobiological model of the generation of the dream perceptual environment. Nonspecific activation of loci in the rhinal cortices back propagates to afferent sites, resulting in a mnemonic parameterization of multimodal and higher order sensory cerebral cortices. Forward propagation into the hippocampal system is presumed to be attenuated by altered brainstem neurotransmission.

Global Workspace and Dreaming 35

long-term memory by “binding” (Cohen & Eichenbaum, 1993) or correlating connection weights in widely dispa-rate areas of cerebral cortex (Halgren, 1984).

What kind of information would be expected to be contained in a “compressed” entorhinal representation? The rhinal cortices, as we reviewed in section 4.4, re-ceive massive input of highly processed sensory infor-mation. Hippocampal function has long been associated with “place cells” and spatial maps in the rat (O’Keefe, 1979, 1990), and spatial learning deficits are associated with hippocampal system damage in monkeys (Rupniak & Gaffan, 1987) and humans (Petrides, 1985). A recent report by Epstein and Kanwisher (1998) has identified in humans a region they term the “parahippocampal place area” which appears to encode the geometry of the local environment during explicit storage, and is most strongly activated when subjects must memorize the placement of objects within an organized environment setting. The PRC has been shown to play an important role in visual object recognition memory (Brown & Xiang, 1998; Murray & Bussey, 1999). Ablation of PRC and PHC produces both visual and tactile memory im-pairments (Suzuki, Zola-Morgan, Squire, & Amaral, 1993). Given the convergence of supramodal inputs spe-cifically to the entorhinal cortex, it would be expected to represent the behavioral salience of sensory environ-ments, in terms of affective value, motor behavioral ac-tivities and salience to homeostatic drives (Suzuki, 1996).

In some sense, therefore, the information “com-pressed” by the rhinal cortices is a distilled representa-tion of the “parameters” required to generate perceptual environments, as well as the patterns of relationship amongst these parameters. We suggest that, within the scope of autoassociative models of the hippocampus, the backpropagation of information from the rhinal cortices to their afferent sites can be construed as a decompres-sion operation. Rhinal cortical representations are “de-compressed” when backpropagation of representations both imparts parameterization and reinstates patterns of relationship (e.g. concurrent activation) to afferent sites. The result is a surrogate sensory input into the relevant thalamocortical loops of multimodal and higher order unimodal sensory cortex (Figure 9).

Following the tenets of the activation-synthesis model, we presume the rhinal cortices are nonspecifi-cally, that is, randomly, activated by cholinergic neuro-transmission. Random activation elicits an arbitrary output from the rhinal cortices, which then backpropa-

gates to their afferent sites. It is critical to grasp the im-plication of random activation of MTLM sites. The MTMS is a system that, by its very nature and function, has evolved to represent any arbitrary set of inputs as structured compressions, as these are prepared for long-term consolidation under normal waking conditions. During waking, MTMS activity is determined by ever-changing sensory and cognitive input. During sleep, this activity is determined by random extrinsic activation. Either source of activating input is, from the MTMS’s point of view, arbitrary to begin with. Running this sys-tem backwards with random input should lead to the “decompression” of a highly organized, but completely arbitrary, representation within MTMS afferent sites. Therefore, given current views of MTMS function, it is not implausible to postulate that a random input could produce a highly organized output to serve as parame-terization for the dream PE (step 1, Figure 10). The ran-dom activation of the MTMS will output any arbitrary set of parameters to afferent sensory cortices, leading to a dream perceptual environment that, although highly structured, is arbitrary.

The phenomenological result is the arbitrary genera-tion of a dream perceptual environment in the dreamer’s consciousness. The WILD experience described in sec-tion 2, and other such descriptions (LaBerge & DeGra-cia, 1999), may actually be first-hand descriptions of the parameterization of the perceptual generator by a back-propagated “decompression” from the rhinal cortices. Likewise, hypnagogic hallucinations may represent ex-actly the same phenomena. Through these phenome-nologies, we can see that the perceptual parameteriza-tion process can occur in stages, first as initial “blips” of perceptual imagery, which eventually cascade and dominate the perceptual generator as full-scale percep-tual environments.

Within the scope of our model, Hunt’s typology of dreaming takes on a new significance. The categories of dreaming he has identified may represent functionally distinct pathways operating in the human declarative memory system. Hunt’s dream typology suggests that different anatomical regions may serve as primary loci for providing parameterization to the dream PE (Figure 8). The four declarative subtypes of dreams derive from complex patterns of relationship produced by the MTMS. Nightmares would be expected to involve MTML-amygdala representations (Suzuki, 1996). Dreams of verbal metaphor could involve representa-tions derived from language production centers, such as

36 DeGracia

Wernicke’s area, that feed into the MTML system (Alkire, Haier, Fallon, & Cahill, 1998; Grasby, Frith, Friston, Frackowiak & Dolan, 1993). What’s more, his typology indicates that some forms of dreaming are less dependent on the MTMS than others such as dreams of verbal metaphor and PG-structured dreams. In these cases, the parameterization comes from regions up-stream of the MTMS. The implication, as alluded to by Antrobus (1993), is that the study of dreaming may shed light on the organization of the human mnemonic sys-tem, for which many methods used in animal research cannot be applied. This view may also explain, at least in part, why such disparate patterns of cortical activity have been reported from neuroimaging studies.

Although our thinking relies on the activation-synthesis model to provide the “impetus” for dream gen-eration in terms of nonspecific activation of the rhinal cortices, it is consistent with Solms’ model. Both of Solms’ “required” brain regions are afferent to the MTMS. In broad terms, the frontal and the parietal re-gions he identified as essential to dreaming correspond roughly to our concepts of cognitive and perceptual gen-erators, respectively. Solms’ “essential” regions may well be those involved in generating the conscious as-pects of dreams, and hence their loss would abolish dream consciousness. Our mechanism would be the unconscious precursor standing upstream to the genera-tion of the dream PE.

What would we expect of the dreams of patients with MTMS damage? Such damage results in antero-grade amnesia during waking (Cohen & Eichenbaum, 1993) but not gross alterations of consciousness per se. We would not expect abatement of dreaming with MTLM damage because these regions are not responsi-ble for generating the conscious output of dreams. Our model suggests that damage to the MTMS should result in dreams of reduced complexity, and abatement of de-clarative memory subtypes of dreams as a function of the extent of MTMS damage. Lesions producing antero-grade amnesia do in fact alter dreaming. Anterograde amnesia is, not surprisingly, associated with poor dream recall (Greenberg, Pearlman, Brooks, Mayer & Hart-mann, 1968; Talland, 1965). Cathala and colleagues (1983) reported altered dream content in amnesic pa-tients, and Torda (1969) reported reduced narrative complexity and emotionality with patients suffering pos-tencephalitic amnesia. Again, given the complexity of interpreting brain lesion data in humans, we cautiously take these observations as support for our model.

Our model is inspired primarily by the relevance of recent findings on the MTML system to the subjective phenomenology of sleep experiences. However, the precedence for our model dates to the 19th century dis-covery by Hughlings Jackson that medial temporal lobe seizures can result in a “dreamy state,” a component of which is vivid “memory-like” hallucinations (Bancaud, Brunet-Bourgin, Chauvel, & Halgren, 1994). Wilder Penfield amplified upon this finding by generating simi-lar “waking dreams” through evoked electrical stimula-tion of the superior temporal gyrus (Penfield & Perot, 1963). More recent studies have investigated the roles of the hippocampal formation and amygdala (Halgren & Wilson, 1985) and medial temporal lobe (Gloor, 1990) in the generation of the “dreamy state.” The similarities between spontaneous epileptic seizures, dreams, and responses to electrical brain stimulation in patients with temporal lobe epilepsy has been described (Ferguson, Rayport, Gardner, Kass, Weiner, & Reiser, 1969). Hence, the mechanism we suggest above may not be confined exclusively to sleep, but may play a more gen-eral role in the genesis of hallucinations during both waking and sleep. DREAMING AS A SELF-CONTAINED INFORMA-TION LOOP

Once the dream PE is generated, a further sequence of events serves to complete the dream generation proc-ess. We imagine, like Bosinelli (1995), that a dream is a self-contained information loop.

Figure 10 depicts our model of the dream produc-tion processes in terms of GW concepts. In figure 10, A1 and A2 refer to hypothesized sources of independent activation for the cognitive and perceptual generators, expressing the dreamer and the dream PE, respectively. Arrow 1 is the backpropagation, from the MTMS to the perceptual generator, of randomly generated patterns of relationship between sensory parameters as discussed above. After this point, however, randomness no longer plays a role, and all of the subsequent events are highly structured by the mnemonic content of the brain. The output of the perceptual generator is the perceptual con-sciousness of the dreamer (step 2, Figure 10). There is next a conscious transaction of information from the perceptual environment to the conscious cognition of the dreamer (step 3, Figure 10).

Stated simply, the dreamer consciously compre-hends the perceptual environment. Following Baars’ notion that information in the global workspace recruits

Global Workspace and Dreaming 37

unconscious contextual information, the conscious com-prehension of the dream environment recruits a contex-tual hierarchy (step 4, Figure 10), which immediately frames the conscious cognition of the dreamer (step 5, Figure 10). The fidelity of this response is a function of the degree of activation of the cognitive generator. Less activation means the dreamer’s response is a degraded form of a potential waking response to the same situa-tion, more activation means the dreamer will more closely approximate the responses of the waking per-sonality.

The perceptual recruitment of a cognitive context to frame the dreamer’s conscious cognition speaks to the notion that dreams possess a formal narrative construc-tion (Cipolli & Poll, 1992; Foulkes, 1978; Kuiken, Neil-son, Thomas, & McTaggart, 1983). Our model, along with other lines of evidence (discussed in Hunt, 1989), suggest that dreams are not formally similar to narrative constructions. The appearance that dreams are “stories,” “plots,” or “narrative” is the result, we suggest, of tem-poral elements in: (1) the perceptual generation process and, (2) in the cognitive context hierarchies elicited by the dream PE. Activation of cognitive contexts will elicit goal-directed behaviors, semantic frameworks, and other mnemonic structures containing substantial tempo-ral dynamics. The temporal unfolding of “routines” or “programs” contained in elicited contexts will create the appearance of a “plot” or “narrative.” Thus, the ostensi-ble narratives of dreams are the time-course of expres-sion of mnemonic elements in the elicited context. That is to say, the temporal unfolding of dream events in terms of the dreamer’s conscious cognition will not re-flect abstract properties of narrative, but will reflect the innate temporal dynamics of elicited contexts, which may be degraded to a greater or lesser extent relative to their waking expression.

Once the sequence of events depicted in Figure 8 is started, the dream becomes a self-contained information loop at both conscious and unconscious levels of infor-mation transfer. Taken together, the conscious output of the perceptual and cognitive generators can be construed as the “manifest” dream. There is also an indirect, un-

conscious interaction from the cognitive generator to the perceptual generator, via the back propagation pathway outlined above. As can be seen in Figure 10, this is actu-ally an unconscious and indirect interaction between different levels of mnemonic information in the cogni-tive generator. In some sense, we could consider this unconscious interaction of information as the dream’s “latent” content. This cycle will continue until either a discreet perceptual discontinuity causes a change in the perceptual environment, and hence begins the sequence anew, or until the generators can no longer support this activity due to decreased activation of the generators, or awakening. LUCID DREAMING AND THE DREAM GENERA-TION PROCESS

Our model of dreaming also speaks to the propensity of lucid dreams to either fade or revert to nonlucid dreams. During lucid dreams, the cognitive generator now bears an increased computing load to support the lucid dream context and output conscious cognition approaching that of the waking personality. The presence of the lucid dream context appears to alter a spontaneous equilib-rium of information transfer from the cognitive to the perceptual generator

If cognitive computing resources are withdrawn from the perceptual generator and used instead to sup-port the lucid dreamer’s conscious cognition, then there is potentially less structuring influence on the perceptual generator (this information flow pathway is labeled in Figure 7). Such a mechanism - a cognitive “loading” - may play some role in the appearance of minimal and surreal environments, which clearly represent perceptual environments of decreased structure. At its farthest ex-treme, too much diversion of cognitive generator com-puting resources away from the perceptual generator could result in fading of the lucid dream and awakening. In this case, the lucid dreamer draws so heavily on cog-nitive computing resources in fulfilling their intention to become the waking personality that they do so by wak-ing up!

38 DeGracia

Interestingly, cognitive “loading” and fading do not appear to occur in reverse. That is, with increasing per-ceptual environment structure, implying “adequate” parameterization of the perceptual generator, there is not an automatic nor incremental decrease in the lucidity of the dreamer. However, with the onset of a fully struc-tured dream perceptual environment, the “relentlessly unfolding” nature of the spontaneous dream process at-

tempts to decrease cognitive generator activity by caus-ing the lucid dream to revert to a nonlucid dream via the competition mechanism discussed in section 7.2.

Hence there appears to be a type of equilibrium as-sociated with nonlucid dreaming: a balance of informa-tion transfer between the perceptual and cognitive gen-erators. Lucid dreaming is a perturbation on this equi-librium, shifting the balance towards cognitive process-

CC

PE

GW

CG PG

A2

2

3

4

5

1

A2

A1

DMS

Fig. 10. Hypothesized sequence for the generation of dreams based on the MTMS operating in reverse. A1 and A2 represent extrinsic sources of activation into the cognitive (CG) and perceptual (PG) generators, respectively. Activation by A1 in the absence of A2 gives rise to conscious cognition (CC) during sleep devoid of a perceptual environment. With the concurrent activation of A1 and A2, the following sequence of information flow generates dreaming: (1) random activation (by A2) of the declarative memory system (DMS) elicits an arbitrary output of unconscious mnemonic information that, because of a posited attenuation of the forward hippocampal circuit, backpropagates to sites afferent to MTMS. This information act as a surrogate “sensory input” to multimodal cortex, providing structuring (parameterization) input that gives rise to the dream perceptual environment; this is an unconscious information flow. (2) The perceptual generator outputs consciousness of the perceptual environment (PE). (3) The perceptual environment is consciously comprehended by the dreamer; a conscious information transaction from the perceptual to the cognitive generator, (4) The conscious cognition of the perceptual envi-ronment recruits unconscious cognitive contexts which come to (5) frame the dreamer’s conscious cognition during the dream. The dream then becomes a feedback cycle at conscious and unconscious levels: conscious between the dreamer’s overt perception and cognition, and unconscious between the input to the perceptual generator and the context framing the dreamer’s conscious cognition. The net result of these information interactions is a random, yet highly structured, recombi-nation of the elements of mnemonic information, which we term “mental recombination.” This cycle will continue until fluctuations in activation alter the input to the perceptual generator, creating a perceptual discontinuity, and beginning the cycle anew. The cycle ends when either (1) activation levels decrease below that needed to support this activity, or (2) the person wakes up and the forward flow of cerebral information reasserts itself due to sensory input.

Global Workspace and Dreaming 39

ing. This disequilibrium attempts to resolve by either (1) fading of the lucid dream, or (2) reversion to a nonlucid dream. This interplay of the lucid dream con-text with perceptual structuring and the spontaneous dream generation process is depicted in Figure 11.

Whether cognitive generator “loading” or competi-tion will dominate a given lucid dream appears to be a function of global brain activation. If global activation is below a critical threshold, then cognitive generator load-ing occurs along with consequent loss of perceptual richness and content. It is as if there is not enough acti-vation to simultaneously support the lucid dream context and a typical dream environment. This may be the un-derlying cause of sensations of dream instability, as well

as the onset of surreal and minimal perceptual environ-ments. Above a critical threshold, the brain becomes activated enough to support both the lucid dream con-text and a typical dream environment, in which case the competition scenario can occur. Therefore, our model predicts that general levels of brain activation (as as-sessed by EEG power analysis or functional neuroimag-ing) will correlate with either “cognitive loading” or competition scenarios, being lower in the former case. Without the excess load imposed by the lucid dream context, this whole series of events does not come into play in nonlucid dreams, with the exception of “embry-onic” dream lucidity, which typically leads to awaken-ing, suggesting a cognitive loading/fading scenario.

CG-PG “equilibrium”

surr

eal e

nviro

nmen

ts

min

imal

env

ironm

ents

awak

enin

g

typi

cal e

nviro

nmen

ts

attenuation of CG → PG structuring via CG “loading”

competition from dream generation process

reve

rsio

n to

no

nluc

id d

ream

Fig. 11. The interaction of the lucid dream context with the spontaneous dream process. The lucid dream context in-creases the computing demands on the cognitive generator (CG), which upsets a conjectured information equilibrium between the cognitive and perceptual generators (PG) during spontaneous nonlucid dreams. To the left, increasing cognitive computing demands divert unconscious computing activity used to structure the dream environment away from the perceptual generator and into conscious cognition (the “loading effect”). The result is decreased structure in the perceptual environment. There appear to be two global “attractor states” in perceptual processing, giving rise to surreal and minimal perceptual environments, before the loading effect is so great that lucid dreaming fading followed by awakening occurs. To the right, when activation levels can sustain both the lucid dream context, and the structuring of the perceptual environment, the spontaneous dream process still seeks to achieve equilibrium by attempting to dis-place the lucid dream context via cognitive competion, and thereby lower cognitive computing demands. This model suggests that there is some critical level of brain activation that serves as the bifurcation point between the left and right scenarios.

40 DeGracia

MENTAL RECOMBINATION

A key implication of our view of dream generation derives from contrasting the contextual structuring of dreams to that of waking. Perceptual structuring is, ac-cording to our model, essentially random; dream envi-ronments simply may or may not resemble familiar wak-ing environments. Perceptual discontinuities of the dream environment are also random (Kahn & Hobson, 1993; Hobson & McCarley, 1977). The cognitive con-textual structuring of the nonlucid dreamer’s conscious-ness is also different from the waking personality for two reasons. First, because the dreamer’s cognition is framed by a degraded version of waking contexts that may form incongruous fusions (Antrobus, 1986), and secondly because whatever cognitive context is evoked occurs in a perceptual environment that may or may not have anything to do with waking experience. Further, all of this occurs within the amnesia of sleep. The net re-sult, in the average case, is Rechtschaffen’s “single-mindedness and isolation of dreams”, even if there is significant resemblance to waking life in any given dream. What we suggest is happening in dreams is that unconscious elements of the waking personality (e.g. the information parameterizing the perceptual environment and the evoked cognitive contexts) are recombined in novel ways within consciousness.

Recall the discussion in section 4.6, that during waking, Baars’ GW model indicates that the function of consciousness is a blend of adapting to novelty and forming long-term contextual structures. The trend of waking conscious function is toward the accumulation of contexts that manifest as habits and routines. As we age and mature, we encounter less novelty and increase our repertoire of habits and expectations. Contrast this now with nonlucid dreams, where there is almost no contribution to consciously accessible long-term mem-ory, and where we see a “hyperassociative” mixing of cognitive and perceptual mnemonic elements. It is clear that dream consciousness does not possess the same function as waking consciousness. Is there anything we can say about putative functions of consciousness in the dream state?

The contrast of consciousness across waking and dreaming suggests that dreams are a form of biological information processing that function analogously to the cellular process of genetic recombination. In genetic recombination, stretches of DNA information exchange position along the length of the chromosome, increasing

the diversity of genetic information (Smith & Nicolas, 1998). Increased genetic diversity enhances adaptability - the ability of biological information structures to alter, and therefore be able respond to unpredictable changes in the environment (Conrad, 1983). At a functional level, genetic recombination is random (Dobzhansky, Ayala, Stebbins, & Valentine, 1977). However, the in-formation that is randomly exchanged is itself highly structured. We suggest dreams serve a similar biocom-putational adaptational role. The generative basis for dreaming within our scheme is that of Hobson & McCarley (1977) and is essentially random. However, random activation of the MTMS leads to the formation of highly structured, albeit arbitrary, perceptual envi-ronments which elicit structured cognitive behavior from the dreamer. The result is a recombination of cerebral information structures during the course of the dream. We refer to this process as mental recombina-tion, by analogy with genetic recombination.

How can mental recombination be functional within the economy of the nervous system? Following a line of thinking that dreams possess adaptive psychological function (Cohen, 1979; Greenberg & Pearlman, 1974; Palambo, 1978; Rotenberg, 1992a, 1992b, 1996), we suggest that dreaming, as mental recombination, serves as a counterpoint to the waking tendency towards con-text (e.g. habit, rote, expectation) formation. Dreaming serves to “loosen” or “mix up” the information con-tained in the brain. Unlike genetic recombination - which is permanent - dreaming, as mental recombina-tion, occurs on a temporary basis. Because of the am-nesia associated with dreaming, dreams typically have no immediate, consciously-accessible effect on the structure of the waking personality. In this sense, men-tal recombination is akin to, for lack of a better term, a type of “novelty priming.” The brief temporal span of dreams is similar to how priming is a temporary, but not a long-term, modification of mnemonic processes (Flowers, 1990; Schacter & Buckner, 1998). In dreams, the brain is briefly exposed to new activity patterns, leaving in most instances only unconscious traces, but not consciously accessible effects. A single dream (re-combination event) is unlikely to have any substantial impact on the waking personality; only very rarely do dreams enter waking consciousness and have a substan-tial long-term impact (Hunt, 1989). It would be ex-pected that the contribution of dreaming to the structure of waking personality is to enhance the flexibility of waking consciousness through numerous “novelty prim-

Global Workspace and Dreaming 41

ing” events which “prime” alternative pathways of neu-ral activation, pathways different from those laid down and reinforced during waking. The increased flexibility of waking consciousness would be expected to be a long-term consequence of dreaming. Our idea of mental recombination falls into a class 2 interpretation of the functional significance of dreams (Fishbein, 1981).

To put the idea that dreams are random recombina-tions of specific levels of neural information into per-spective, we need to be cognizant of the fact that the “constructive utilization of randomness” is a general feature of biological information processing (Conrad, 1985). Random mutations and recombinations of genetic information are the fodder from which the evolution of species arise. Heat (e.g. random molecular motion), a notorious source of noise to the engineer, is the basis of diffusional forces which allow biological chemical reac-tions to operate (Conrad, 1985). The release of synaptic vesicles (Katz & Miledi, 1972), the opening and closing of ion channels (Liebovitch & Tóth, 1991), and single neuron pulse trains (Freeman, 1996) are stochastic proc-esses, the macroscopic results of which are the electrical and neurocomputational properties of the brain. Steven Thaler’s Creativity Engines are neural network systems that display creative behavior; this effect is achieved by introducing a precise degree of random perturbation into the neural networks (Thaler, 1996a, 1996b). Thaler’s model indicates there is an optimal degree of random-ness that is “good” for a biological system: too much randomness and you get a mess, too little, and things do not change enough to allow “creativity” to emerge. Dreams may play a similar optimization role in the bal-ance between flexibility and stability in cerebral infor-mation structures.

Stated simply, within biological systems random-ness can be a good thing. Randomness serves as the necessary basis for ostensibly creative actions by bio-logical systems which, in fact, are not intentionally crea-tive but simply spontaneously emerge. More precisely, randomness is the necessary basis for the flexibility that allows biological systems to adapt to unpredictable changes in their environment. The brain, as a product of evolution by natural selection, would be fully expected to utilize the same types of adaptive mechanisms as any other biological information system. Hence, Hobson and McCarley’s infamous notion that dreams are ran-dom has been misconstrued (such as the discussion in Moffitt, Kramer, & Hoffmann, 1993). This random-ness is not meaningless, but is an essential ingredient

that allows dreams to serve a role in the adaptive bio-computational properties of the brain. This, we suggest is precisely why dreaming evolved: it imparts an adap-tive advantage over a hypothetical nondreaming brain which, according to the GW model, would become en-crusted with habits of thought and action over time as context formation would overwhelm the flexibility as-pects of consciousness. Because we dream, our minds experience a flexibility that provides the fodder for our adaptive, creative and new responses during waking (Greenberg & Pearlman, 1974). Since the essential func-tion of dreaming as mental recombination is to simply recombine mnemonic information on a temporary basis, there is, at least in principle, no intrinsic need to remem-ber our dreams overtly; the unconscious “novelty prim-ing” effects are sufficient for dreaming to exert its func-tion. In fact, remembering all of our dreams would waste neural computing resources and would be counter-adaptive, as some authors have pointed out (e.g. Kava-nau, 1996). However, there are important caveats to this point we discuss below.

COMPARISON OF MENTAL RECOMBINATION TO OTHER MODELS OF DREAM FUNCTION

We would like to end by contrasting a few contem-porary ideas of dream function with our concept of men-tal recombination.

Hunt (1989) has presented the idea that dreams serve a “recombinatory” function, a notion also ex-pressed in Mancia (1995). The result of recombination in dreams, according to Hunt (1989), is to integrate newly learned information into stores of older informa-tion, or that dreaming plays a role in short-term to long-term explicit memory consolidation. In contrast, we envision dreams as a literal “mixing up” of cerebral in-formation solely for the purpose of increasing the flexi-bility of neural circuitry. Hunt’s idea predicts that we should be able to consciously recall new information better after dreaming. Our idea predicts that dreaming should make waking cognitive activities more flexible (e.g. adaptable) than if dreaming was not present, but no single dream, on average, will have an overt affect on waking conscious operations. We suggest that the am-biguity of the results from sleep deprivation experi-ments on waking learning and memory performance (Rosa & Bonnet, 1985; Rotenberg, 1992b; Vogel, 1975) indicates that dream deprivation does not have an im-mediate, readily discernible effect on waking explicit recall. Increased REM duration following unprepared

42 DeGracia

learning seems to support some role of REM in the con-solidation process (Rotenberg, 1996). However, Ro-tenberg’s arguments that increased REM following un-prepared learning is due to anxiety reduction mecha-nisms not necessarily related to memory consolidation cannot be lightly dismissed (Rotenberg & Arshavsky 1979; Rotenberg 1992a, 1992b).

Nonetheless, our and Hunt’s notions are not mutu-ally exclusive. First, the addition of any newly learned waking material can serve as fodder in the random re-combination process, which would also serve to rein-force and integrate newly learned material. Perhaps more importantly, the impact of “day-residue,” or im-mediate problems of waking life (Beauchemin & Hays, 1996; Cartwright, 1993), can serve as modulating influ-ences on the random basis of dream generation. Waking concerns and problems could readily serve as perturba-tions and bias the random dream generation process to-wards the mnemonic structure of waking concerns and problems, which is the opposite emphasis of Cohen (1979) who saw dreams as generating novel solutions to presleep problems. The random mechanism we postu-lated above is not rigid and fixed, but can be modulated by existent memories. Such a position reconciles the random biological basis of dreaming with the widely recognized psychological significance of dreaming. Per-haps the ultimate example of a mnemonic structure modulating the random dream process is the lucid dream context, where the waking personality directly inter-venes in the spontaneous dream generation process. We will not pursue this line of thought further here. The issue of how the conscious cognition of the waking per-sonality can bias the unconscious structuring of the per-ceptual environment is an important question (e.g. Worsley, 1988; Tholey, 1988), and one that we hope our ideas make more tractable.

Our views also speak to Crick and Mitchenson’s (1983, 1995) concept that dreams are a form of “reverse learning.” Under certain conditions, overlapping inputs to a neural net will result in overlapping outputs. Spe-cific mathematical manipulations of the neural net will cause it to cease overlapping the outputs even when the inputs continue to overlap. This change in the neural net is called “reverse learning.” Crick and Mitchenson have postulated that dreaming plays an analogous role in brain function such that dreams lead to a refinement of waking information structures by erasing “noise” from mnemonic circuits. However, the mechanism of reverse learning is similar to the act of intentional practice,

which refines behavior during waking, and hence plays a role in the tendency towards habit formation during waking. That is to say, reverse learning is probably an important aspect of the neurocomputations underlying procedural memory formation, or Baars’ (1988) adapta-tion cycle, but it probably plays little role, if any, in normative nonlucid dreams. Reverse learning could po-tentially operate in lucid dreams when refinement of the lucid dream context occurs.

In a somewhat similar vein, Kavanau (1996) has suggested that dreams have evolved to allow a “dynamic stabilization” of infrequently used neural circuits, and suggests that this is also the basis for the evolution of the inability to remember our dreams. We point out that this view fails to take into account the well known “use it or lose it” principle in the brain. The waking brain has a mechanism for dealing with infrequently used explicit information: it overwrites it with relevant and frequently used information. On the other hand, procedural memo-ries can be maintained for long periods without waking reinforcement, and there is evidence of the involvement of REM sleep with consolidation of procedural memo-ries (Karni, Tanne, Rubenstein, Askenasy, & Sagi, 1994; Smith & Lapp, 1986; Smith, 1996). Kavanau’s postulated mechanism may play some role in maintain-ing procedural memory during dreaming, and is perhaps related to the consistent fidelity of overlearned activities across waking and dreaming identified above.

Finally, we would like to comment on how lucid dreaming fits into the view of dream function we have outlined here. Reasonable estimates suggest that lucid dreams take up only a very small percentage of an ex-perienced lucid dreamer’s total dream time (LaBerge, Levitan, DeGracia, & Zarcone, 1999). Hence, over the long-term, lucid dreaming would not be expected to in-terfere with dream function as we have construed it here. In spite of being a somewhat fragile balancing act, lucid dreams are remembered well after waking, as opposed to leaving only unconscious “novelty priming” traces in the brain after awakening from nonlucid dreaming. From the point of view that the spontaneous dream process can be biased towards manifesting waking prob-lems or concerns, or even imbalances in the waking per-sonality, lucid dreaming clearly provides a huge accel-eration over nonlucid dreams in bringing these factors to the attention of the waking personality. Hence, lucid dreams can, and should, be used as a form of self-understanding (Malamud, 1988; Tholey, 1988).

Otherwise, the recognition that dreaming is ulti-

Global Workspace and Dreaming 43

mately an expression of unconscious forms of informa-tion into the consciousness of the dreamer, via the type of synesthetic mechanisms discussed by Hunt (1989), or via the model we have developed above, has within it the potential to allow the development of a whole new level of subjective description of conscious experience via lucid dreaming. Specifically, the development of methods and ideas that will allow us to begin to com-prehend the nature of surreal dream environments and other forms of conscious experience with no waking counterpart is strongly inviting (e.g. see Rifat, 1980, 1997). This has been approached in the Tibetan lucid dreaming tradition (Gillespie, 1988). However, the modern lucid dreamer has all the knowledge and discov-eries of modern cognitive neuroscience upon which to draw in ascertaining the nature of things experienced uniquely in the lucid dream state. We strongly encour-age this line of intellectual evolution.

CONCLUSION

Table 3 summarizes the salient similarities and dif-ferences we have identified between conscious and un-conscious processes across waking, nonlucid and lucid dreaming. These ideas have been presented in the spirit of unifying our understanding of the mind-brain system in its entirety. By integrating the phenomenology of sleep experiences into the Global Workspace model, we have developed a framework that we feel provides a means for discussing across-state changes in conscious and unconscious processes. We have explored a num-ber of the implications of this way of thinking, but have by no means exhausted the potential contained therein. Our thinking has been primarily inspired by the phe-nomenology of actual human experiences in the sleep state. Therein lies both its strength and its weakness. Whether or not every detail we discussed above turns out to be accurate, our ultimate intention is to stimulate a more unified dialogue, one which seeks to encompass

Waking Nonlucid Dreaming Lucid Dreaming PERCEPTUAL PG parameterization source

sensory mnemonic mnemonic

Environment types typical typical typical, minimal and surreal

Bizarreness none: consistent present present Conscious sensations of perceptual stabil-ity/instability

not present not present present; from stable to unstable; when highly unstable, dream fades

COGNITIVE Cognitive Fidelity relatively constant widely variable variable Affect baseline exaggerated similar to waking Personality structure intact displays degrees of

“graceful degradation” intact, with possible cognitive degradation

Long-term episodic memory formation

intact weak; may recall last dream prior to waking

intact; somewhat below waking levels

Contextual structure framing conscious op-eration

waking personality highly variable; isolated from waking personality

waking personality via lucid dream context

Function of conscious processes

recognize novelty; adapt to regularities

mental recombination: generate conscious nov-elty via recombination of declarative and per-ceptual contexts

recognize novelty; adapt to regularities; generate conscious novelty via recombination of de-clarative and perceptual contexts

Table 3: Comparison of waking, nonlucid and lucid dreams in terms of perceptual and cognitive gen-

erator activities.

44 DeGracia

the multitude of facets of human experience. We hope at least in this we have succeeded.

Acknowledgements

The author would like to thank John Collins and two anonymous reviewers for their helpful and insightful comments.

References Aaronson, B., & Osmond, H. (Eds.). (1970). Psychedelics. New York: Doubleday & Co. Inc.

Alkire, M.T., Haier, R.J., Fallon, J.H., & Cahill L. (1998). Hippocampal, but not amygdala, activity at encoding correlates with long-term, free recall of nonemotional information. Proceedings of the National Acad-emy of Sciences of the United States of America 95, 14506-10.

Amaral, D.G., Insausti, R., & Cowan, W.M. (1987). The entorhinal cor-tex of the monkey: I. Cytoarchitectonic organization. Journal of Com-parative Neurology 264, 326-355.

Antrobus, J. (1986). Dreaming: Cortical activation and perceptual thresholds. The Journal of Mind and Behavior 7, 193-212.

Antrobus, J. (1987). Cortical hemisphere asymmetry and sleep menta-tion. Psychology Review 94, 359-368.

Antrobus, J. (1991). Dreaming: Cognitive processes during cortical acti-vation and high afferent thresholds. Psychological Review 98, 96-121.

Antrobus, J. (1993). Dreaming, could we do without it? In A. Moffitt, M. Kramer & R. Hoffmann (Eds), The functions of dreaming. Albany, NY: State University of New York Press. pp. 549-558.

Antrobus, J., Hartwig, P., Rosa, D., Reinsel, R. & Fein, G. (1987). Brightness and clarity of REM and NREM imagery: photo response scale. Sleep Research 16, 240.

Antrobus, J., Kondo, T., Reinsel, R., & Fein, G. (1995). Dreaming in the late morning: summation of REM and diurnal cortical activation. Con-sciousness and Cognition 4, 275-299.

Baars, B.J. (1988). A cognitive theory of consciousness. Cambridge: Cambridge University Press.

Baars, B.J. (1993). How does a serial, integrated and very limited stream of consciousness emerge from a nervous system that is mostly uncon-scious, distributed, parallel and of enormous capacity? Ciba Foundation Symposium 174, 282-303.

Baars, B.J. (1996a). When are images conscious? The curious disconnec-tion between imagery and consciousness in the scientific literature. Con-sciousness and Cognition 5, 261–264.

Baars, B.J. (1996b). Understanding subjectivity: global workspace theory and the resurrection of the observing self. Journal of Consciousness Studies 3, 211-216.

Baars, B.J. (1997a). In the theater of consciousness. New York: Oxford University Press.

Baars, B.J. (1997b). Some essential differences between consciousness and attention, perception, and working memory. Consciousness and Cognition 6, 363–371.

Bancaud, J., Brunet-Bourgin, F., Chauvel, P., & Halgren, E. (1994). Anatomical origin of deja vu and vivid 'memories' in human temporal lobe epilepsy. Brain 117, 71-90.

Beauchemin, K.M., & Hays, P. (1996). Dreaming away depression: The role of REM sleep and dreaming in affective disorders. Journal of Affec-tive Disorders 41, 125-133.

Benson, D.F. (1994). The neurology of thinking. New York: Oxford University Press.

Berridge, C.W., Page, M.E., Valentino, R.J., & Foote, S.L. (1993). Ef-fects of locus coeruleus inactivation on electroencephalographic activity in neocortex and hippocampus. Neuroscience 55, 381-393.

Bertenthal, B. (1996). Origins and early development of perception, action, and representation. Annual Review of Psychology 47, 431–459.

Black, J.E., & Greenough, U.T. (1986). Induction of pattern in neural structure by experience: Implications for cognitive development. In M.E. Lamb, A.L. Brown, & B. Rogoff (Eds.), Advances in developmental psychology, vol. 4, pp. 1-5. Hillsdale, New Jersey: Lawrence Erlbaum Associates.

Blackmore, S. (1988). A theory of lucid dreams and OBEs. In J. Gack-enbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain, pp. 373-387. New York: Plenum Press.

Blackmore, S. (1982). Have you ever had an OBE?: The wording of the question. Journal for the Society of Psychical Research 51, 292-301.

Bonin, G.V. & Baily, P. (1947). The neocortex of macaca mulatta. Ur-bana, IL: University of Illinois Press.

Bonnet, M.H. (1986). Auditory thresholds during continuing sleep. Bio-logical Psychology 22, 3-10.

Bonnet, M.H., & Johnson, L.C. (1978). Relationship of arousal threshold to sleep stage distribution and subjective estimates of depth and quality of sleep. Sleep 1, 161-168.

Bosinelli, M. (1995). Mind and consciousness during sleep. Behavioral Brain Research 69, 195-201.

Boss, M. (1958). The analysis of dreams. New York: Philosophical Li-brary.

Braak, H. & Braak, E. (1992). The human entorhinal cortex: normal morphology and lamina-specific pathology in various diseases. Neuro-science Research 15, 6-31.

Braun, A.R., Balkin, T.J., Wesensten, N.J., Gwadry, F., Carson, R.E., Varga, M., Baldwin, P., Belenky, G., & Herscovitch, P. (1998). Dissoci-ated pattern of activity in visual cortices and their projections during human rapid eye movement sleep. Science 279, 91-95.

Braun, A.R., Balkin, T.J., Wesensten, N.J., Carson, R.E., Varga, M., Baldwin, P., Selbie, S., Belensky, G., & Herscovitch, P. (1997). Re-gional cerebral blood flow throughout the sleep-wake cycle An H2

15O PET study. Brain 120, 1173–1197.

Broughton, R. (1982). Human consciousness and sleep/waking rhythms: A review and some neuropsychological considerations. Journal of Clini-cal Neuropsychology 4, 193-218.

Brown, M.W., & Xiang, J.Z. (1998). Recognition memory: neuronal substrates of the judgement of prior occurrence. Progress in Neurobiol-ogy 55, 149-189.

Global Workspace and Dreaming 45

Brown, T.H., & Zador, A.M. (1990). Hippocampus. In G.M. Shepard (Ed.), The Synaptic Organization of the Brain, vol. 3, pp. 346-388. New York, NY: Oxford University Press.

Bullmore, E.T., Rabe-Hesketh, S., Morris, R.G., Williams, S.C., Greg-ory, L., Gray, J.A., & Brammer, M.J. (1996). Functional magnetic reso-nance image analysis of a large-scale neurocognitive network. Neuroi-mage 4,16-33.

Busby, K.A., Mercier, L., & Pivik, R.T. (1994). Ontogenetic variations in auditory arousal threshold during sleep. Psychophysiology 31, 182-188.

Buzsáki, G. (1989). Two-stage model of memory trace formation: a role for “noisy” brain states. Neuroscience 31, 551-570.

Cartwright, R.D. (1993). Who needs their dreams? The usefulness of dreams in psychotherapy. Journal of the American Academy of Psycho-analysis. 21, 539-547.

Cathala, H., Laffont, F., Siksou, M., Esnaut, S., Gilbert, A., Minz, M., Moret-Chalmin, C., Buzarè, M., & Waisbord, P. (1983). Sommeil et rêve chez des patients atteins de lèsions pariètales et frontales. Revue Neu-rologique 139, 497-508.

Cavada, C., & Goldman-Rakic, P.S. (1989). Posterior parietal cortex in rhesus monkey: parcellation of areas based on distinctive limbic and sensory corticocortical connections. The Journal of Comparative Neu-rology 287, 393-421.

Cavallero, C. (1987). Dream sources, associative mechanisms, and tem-poral dimension. Sleep 10, 78-83.

Cavallero, C., Cicogna, P., Natale, V., Occhionero, M., & Zito, A. (1992). Slow wave sleep dreaming. Sleep 15, 562-566.

Cavallero, C., Foulkes, D., Hollifield, M., & Terry, R. (1990). Memory sources of REM and NREM dreams. Sleep 13, 449-455.

Cawte, J. (1984). The 'ordinary' dreams of the Yolngu in Arnhem Land. Australian & New Zealand Journal of Psychiatry 18, 236-243.

Chen, W., Kato, T., Zhu, X.H., Ogawa, S., Tank, D.W., & Ugurbil, K. (1998). Human primary visual cortex and lateral geniculate nucleus acti-vation during visual imagery. Neuroreport 9, 3669-3674.

Chrobak, J.J., & Buzsáki, G. (1998). Operational dynamics in the hippo-campal-entorhinal axis. Neuroscience & Biobehavioral Reviews 22, 303-310.

Cicogna, P., Cavallero, C., & Bosinelli, M. (1991). Cognitive aspects of mental activity during sleep. American Journal of Psychology 104, 413-425.

Cipolli, C., & Poll, D. (1992). Story structure in verbal reports of mental sleep experiences after awakening in REM sleep. Sleep 152, 133-142.

Cline, H.T. (1998).Topographic maps: Developing roles of synaptic plasticity. Current Biology 8, R836- R839.

Cohen, D. H. (1979). Sleep and dreaming: origins, nature, and function. Oxford: Pergamon Press.

Cohen, N.J., & Eichenbaum, H. (1993). Memory, amnesia and the hip-pocampal system. Cambridge, Massachusetts: MIT Press.

Conrad, M. (1983). Adaptability: The significance of variability from molecule to ecosystem. New York: Plenum Press.

Conrad, M. (1985). On design principles for a molecular computer. Communications of the Association for Computing Machinery 28, 464-480.

Covello, E. (1984). Lucid dreaming: A review and experiential study of waking intrusions during stage REM sleep. The Journal of Mind and Behavior 5, 81-98.

Crick, F., & Mitchison, G. (1983). The function of dream sleep. Nature 304, 111-114.

Crick, F., & Mitchison, G. (1995). REM sleep and neural nets. Behav-ioral Brain Research 69, 147-155.

D’esposito, M., & Grossman, M. (1996). The physiological basis of executive function and working memory. The Neuroscientist 2, 345-352.

Damasio, A., & Damasio, H. (1996). Making images and creating sub-jectivity. In R. Llinás & P. Churchland (Eds.) The Mind-Brain contin-uum. MA: MIT Press, pp. 19-28.

DeGracia, D.J. (1997). Les paradigmes de la conscience dans le sommeil. Rêver 1, 26-35.

DeGracia, D.J., & LaBerge, S. (1999). Using global workspace theory to interpret dreams and lucid dreams. In preparation.

Dement, W., & Kleitman, N. (1957). Cyclic variations in EEG during sleep and their relation to eye movements, body motility, and dreaming. Electroencephalography and Clinical Neurophysiology 9, 673-690.

Dobzhansky, T. Ayala, F.J. Stebbins, G.L., & Valentine, J.W. (1977). Evolution. San Fransico: W.H. Freeman & Co.

Dorus, E., Dorus, W., & Rechtschaffen, A. (1971). The incidence of novelty in dreams. Archives of General Psychiatry 25, 364-368.

Eichenbaum, H. (1998). Declarative memory: Insights from cognitive neurobiology. Annual Review of Psychology 49, 547-572.

Eichenbaum, H., & Bunsey, M. (1995). On the binding of associations in memory: Clues from studies on the role of the hippocampal region in paired associative learning. Current Directions in Psychological Sci-ence, 4, 19-23.

Eichenbaum, H., & Otto, T. (1993). Where perception meets memory: Functional coding in the hippocampus. In T. Ono, L.R. Squire, M.E. Raichle, D.I. Perret, & M. Fukuda (Eds.), Brain mechanisms of percep-tion and memory from neuron to behavior, pp. 301-329. New York: Oxford University Press.

Empson, J. (1989). Sleep and dreaming. London: Faber & Faber.

Epstein, A.W. (1979). Effect of certain cerebral hemispheric diseases on dreaming. Biological Psychiatry 14, 77-93.

Epstein, R., & Kanwisher, N. (1998). A cortical representation of the local visual environment. Nature, 392, 598-601.

Erisir, A., Van Horn, S.C., & Sherman, S.M. (1997). Relative numbers of cortical and brainstem inputs to the lateral geniculate nucleus. Pro-ceedings of the National Academy of Sciences of the United States of America 94, 1517-1520.

Farah, M.J. (1994). Neuropsychological inference with an interactive brain: A critique of the "locality" assumption. Behavioral and Brain Sciences 17, 43-104.

Farah, M.J., Peronnet, F., Gonon, M.A., & Girard, M.H. (1988). Electro-physiological evidence for a shared representational medium for visual

46 DeGracia

images and visual percepts. Journal of Experimental Psychology 3, 99–109.

Felleman, D.D., & Van Essen, D.C. (1991). Distributed hierarchical processing is the primate cerebral cortex. Cerebral Cortex 1, 1-47.

Ferguson, S.M., Rayport, M., Gardner, R., Kass, W., Weiner, H., & Reiser, M.F. (1969). Similarities in mental content of psychotic states, spontaneous seizures, dreams, and responses to electrical brain stimula-tion in patients with temporal lobe epilepsy. Psychosomatic Medicine 31, 479-498.

Fishbein, W. (1981). Sleep, dreams and memory. New York: SP Medical and Scientific Books.

Fiss, H. (1983). Toward a clinically relevant experimental psychology of dreaming. Hillside Journal of Clinical Psychiatry 5, 147-159.

Flowers, J.H. (1990). Priming effects in perceptual classification. Per-ceptual Psychophysics 47, 135-148.

Foss, D.J., & Hakes, D.T. (1978). Psycholinguistics: An introduction to the psychology of language. New Jersey: Prentice Hall.

Foulkes, D. (1962). Dream reports from different stages of sleep. Jour-nal of Abnormal Social Psychology, 65, 14-25.

Foulkes, D. (1967). Nonrapid eye movement mentation. In C.D. Clemente (Ed.), Experimental Neurology Supplement 4: Physiological Correlates of Dreaming.

Foulkes, D. (1978). A grammar of dreams. New York: Basic Books.

Foulkes, D. (1982). A cognitive-psychological model of REM dream production. Sleep 5, 169-187.

Foulkes, D. (1991). Dreaming: lucid and non. Lucidity, 10(1-2), 272-274.

Foulkes, D., & Cavallero, C. (1993). Introduction. In C. Cavallero, & D. Foulkes (Eds.), Dreaming as cognition, pp. 1-17. New York: Harvester Wheatsheaf.

Foulkes, D., & Kerr, N.H. (1994). Point of view in nocturnal dreaming. Perceptual and Motor Skills 78, 690.

Freeman, W.J. (1994). Neural networks and chaos. Journal of Theoreti-cal Biology 171, 13-18.

Freeman, W.J. (1996). Random activity at the microscopic neural level in cortex ("noise") sustains and is regulated by low-dimensional dynam-ics of macroscopic cortical activity ("chaos"). International Journal of Neural Systems 7, 473-480.

Freeman, W.J., & Skarda, C.A. (1985). Spatial EEG patterns, non-linear dynamics and perception: the neo-Sherringtonian view. Brain Research 357, 147-175.

Freud, S. (1900). The interpretation of dreams. New York: Avon, 1983.

Gabel, S. (1993). The phenomenology of the self and its objects in wak-ing and dreaming: Implications for a model of dreaming. Journal of the American Academy of Psychoanalysis 21, 339-362.

Gabrieli, J.D.E. (1998). Cognitive neuroscience of human memory. An-nual Review of Psychology 49, 87-115.

Gackenbach, J. (1988). The psychological content of lucid versus nonlu-cid dreams. In J. Gackenbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspectives on lucid dreaming, pp. 181-220. New York: Plenum Press.

Gackenbach, J. (1991a). An estimate of lucid dreaming incidence. Lucid-ity, 10(1-2), 232-233.

Gackenbach, J. (1991b). From lucid dreaming to pure consciousness: a conceptual framework for the OBE, UFO abduction and NDE experi-ences. Lucidity, 10(1-2), 277-307.

Gackenbach, J., Cranson, R., & Alexander, C. (1986). Lucid dreaming, witness dreaming, and the transcendental meditation technique. Lucidity Letter 5, 34-40.

Gaffan, D. (1994). Dissociated effects of perirhinal cortex ablation, fornix transection and amygdalectomy: evidence for multiple memory systems in the primate temporal lobe. Experimental Brain Research 99, 411-422.

Gaillard, J.M. (1985) Neurochemical regulation of the states of alertness. Annals of Clinical Research 17, 175-84.

Gastaut, H. (1969). Hypnosis and pre-sleep patterns. In L. Chertok (Ed.), Psychophysiological mechanisms of hypnosis. Berlin: Springer-Verlag.

Geof, W., Allison T., Shapiro A., & Rosner B. (1966). Cerebral somato-sensory responses evoked during sleep in man. Electroencephalography and Clinical Neurophysiology 21, 1-19.

Gillespie, G. (1989). Lights and lattices and where they are seen. Percep-tual and Motor Skills 68, 487-504.

Gillespie, G. (1983). Memory and reason in lucid dreams: A personal observation. Lucidity Letter, 3, 76-77.

Gillespie, G. (1984). Problems related to experimentation while dream-ing lucidly. Lucidity Letter 3(2&3), 84-89.

Gillespie, G. (1988). “Lucid dreams in Tibetan Buddhism”. In J. Gack-enbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspec-tives on lucid dreaming, pp. 27-36. New York: Plenum Press.

Globus, G.G. (1992). Toward a noncomputational cognitive neurosci-ence. Journal of Cognitive Neuroscience 4, 299-310.

Gloor, P. (1990). Experiential phenomena of temporal lobe epilepsy. Facts and hypotheses. Brain 113, 1673-1694.

Gluck, M.A., & Myers, C.E. (1997). Psychobiological models of hippo-campal function in learning and memory. Annual Review of Psychology 48, 481-514.

Goldman-Rakic, P.S. (1995). Anatomical and functional circuits in pre-frontal cortex of nonhuman primates. In H.H. Jasper, S. Riggio, & P.S. Goldman-Rakic (Eds.), Epilepsy and the functional anatomy of the fron-tal lobe, pp. 51-65. New York: Raven Press Ltd.

Graf, P. & Schacter, D.L. (1985). Implicit and explicit memory for new associations in normal and amnesic patients. Journal of Experimental Psychology: Learning, Memory and Cognition 11, 501-518.

Grasby, P.M., Frith, C.D., Friston, K., Frackowiak, R.S., & Dolan, R.J. (1993). Activation of the human hippocampal formation during auditory-verbal long-term memory function. Neuroscience Letters, 12, 185-188.

Green, C. (1968). Lucid dreams. London: Hamish Hamilton.

Green, C., & McCreery, C. (1975). Apparitions. London: Hamish Hamil-ton.

Green, C., & McCreery, C. (1994). Lucid dreaming: The paradox of consciousness during sleep. London: Routledge.

Global Workspace and Dreaming 47

Greenberg, R., Pearlman, C., Brooks, R., Mayer, R., & Hartmann, E. (1968). Dreaming and Korsakoff’s psychosis. Archives of General Psy-chiatry 18, 203-209.

Greenberg, R., & Pearlman, C. (1974). Cutting the REM nerve: An ap-proach to the adaptive role of REM sleep. Perspectives in Biology and Medicine 17, 513-521.

Greenberg, S.N. (1977). Competition between memory and perceptual tasks involving physically similar stimuli. American Journal of Psychol-ogy 90, 675-687.

Guido, W., & Lu, S. (1995). Cellular basis for the control of retinogeni-culate signal transmission. International Journal of Neurosciences 80, 41-63.

Gulyas, B. (1997). Functional organization of human visual cortical areas. Cerebral Cortex 12, 743-755.

Halgren, E. (1984). Human hippocampal and amygdala recording and stimulation: evidence for a neural net model of recent memory. In L.R. Squire & N. Butters (Eds.), The Neuropsychology of Memory. New York, NY: Guilford Press. Pp. 165-182.

Halgren, E., & Wilson, C.L. (1985). Recall deficits produced by afterdis-charges in the human hippocampal formation and amygdala. Electroe-ncephalography and Clinical Neurophysiology 61, 375-380.

Hall, C., & Van de Castle, R. (1966). The content analysis of dreams. New York: Appelton-Century-Crofts.

Hardcastle, V. (1994). Psychology’s binding problem and possible neurobiological solutions. Journal of Consciousness Studies 1, 66-90.

Harley, C.W. (1987). A role for norepinephrine in arousal, emotion and learning?: limbic modulation by norepinephrine and the Kety hypothesis. Progress in Neuropsychopharmacology and Biological Psychiatry 11, 419-458.

Harley, C.W. (1991). Noradrenergic and locus coeruleus modulation of the perforant path-evoked potential in rat dentate gyrus supports a role for the locus coeruleus in attentional and memorial processes. Progress in Brain Research 88, 307-321.

Harley, C.W. (1998). Noradrenergic long-term potentiation in the dentate gyrus. Advanced in Pharmacology 42, 952-956.

Hartmann, E. (1973). The function of sleep. New Haven: Yale University Press.

Hartmann, E. (1982). From the biology of dreaming to the biology of the mind. Psychoanalytic Study of the Child 37, 303-335.

Hasselmo, M.E. (1995). Neuromodulation and cortical function: model-ing the physiological basis of behavior. Behavioral Brain Research 67, 1-27.

Hearne, M.K.T. (1981). A ‘light switch’ phenomena in lucid dreams. Journal of Mental Imagery 5, 97-100.

Heather-Greener, G.Q., Comstock, D., & Joyce, R. (1996). An investiga-tion of the manifest dream content associated with migraine headaches: A study of the dreams that precede nocturnal migraines. Psychotherapy & Psychosomatics 65, 216-221.

Hess, C.W., Mills, K.R., Murray, N.M., & Schriefer, T.N. (1987). Excit-ability of the human motor cortex is enhanced during REM sleep. Neuro-science Letters 10, 47-52.

Hishikawa, Y., & Shimizu, T. (1995). Physiology of REM sleep, cata-plexy, and sleep paralysis. Advances in Neurology 67, 245-271.

Hobson, J.A. (1988). The dreaming brain. New York: Basic Books.

Hobson, J.A., & McCarley, R.W. (1977). The brain as a dream state generator: An activation-synthesis hypothesis of the dream process. American Journal of Psychiatry 134, 1335-1348.

Hobson, J.A., & Schmajuk, N.A. (1988). Brain state and plasticity: an integration of the reciprocal interaction model of sleep cycle oscillation with attentional models of hippocampal function. Archives of Italian Biology 126, 209-224.

Hobson, J.A., & Steriade, M. (1986). Neuronal basis of behavior state control. In V.B. Mountcastle, F.E. Bloom, & S.R. Geiger (Eds.), Hand-book of Physiology; Section I The Nervous System, vol. 4. Bethesda, Maryland: American Physiological Society.

Hobson, J. A., & Stickgold, R. (1995). The conscious state paradigm: A neurocognitive approach to waking, sleeping, and dreaming. In M. Gaz-zaniga (Ed.), The Cognitive Neurosciences. Cambridge, Massachusetts: MIT Press.

Hobson, J.A., Pace-Schott, E.F., Stickgold, R., & Kahn, D. (1998a). To dream or not to dream? Relevant data from new neuroimaging and elec-trophysiological studies. Current Opinions in Neurobiology 8, 239-244.

Hobson, J.A., Stickgold, R., & Pace-Schott, E.F. (1998b). The neuropsy-chology of REM sleep dreaming. Neuroreport 9, R1-R14.

Hoover, D.B., & Baisden, R.H. (1980). Localization of putative cho-linergic neurons innervating the anteroventral thalamus. Brain Research Bulletin 5, 519-524.

Hopfield, J.J. (1982). Neural networks and physical systems with emer-gent collective computational abilities. Proceedings of the National Academy of Sciences of the United States of America 81, 3088-3092.

Howard, R.J., Fytche, D.H., Barnes, J., McKeefry, D., Ha, Y., Woodruff, P.W., Bullmore, E.T., Simmons, A., Williams, S.C., David, A.S., & Brammer, M. (1998). The functional anatomy of imagining and perceiv-ing colour. Neuroreport 9, 1019-1023.

Hunt, H.T. (1989). The multiplicity of dreams. New Haven: Yale Univer-sity Press.

Hunt, H.T., Ogilvie, R., Belicki, K., Beliki, D., & Atalik, E. (1982). Forms of dreaming. Perceptual & Motor Skills 54, 559-633.

Insausti, R., Amaral, D.G., & Cowan, W.M. (1987). The entorhinal cor-tex of the monkey: II. Cortical afferents. Journal of Comparative Neu-rology 264, 356-395.

Izquierdo, I., & Medina, J.H. (1995). Correlation between the pharma-cology of long-term potentiation and the pharmacology of memory. Neurobiology of Learning and Memory 63, 19-32.

Izumi, Y., & Zorumski, C.F. (1999). Norepinephrine promotes long-term potentiation in the adult rat hippocampus in vitro. Synapse 31, 196-202.

Johnson, M.K., O’Conner, M., & Cantor, J. (1997). Confabulation, memory deficits and frontal dysfunction. Brain and Cognition 34, 189-206.

Jones, B.E. (1991). Paradoxical sleep and its chemical/structural sub-strates in the brain. Neuroscience 40, 637-656.

48 DeGracia

Jones, B.E., & Cuello, A.C. (1989). Afferents to the basal forebrain cho-linergic cell area from pontomesencephalic--catecholamine, serotonin, and acetylcholine--neurons. Neuroscience 31, 37-61.

Jung, C.G. (1961). Memories, dreams and reflections. New York: Vin-tage Books, 1989.

Kaas, J.H. (1997). Topographic maps are fundamental to sensory proc-essing. Brain Research Bulletin 44,107-112.

Kahan, T., & LaBerge, S. (1994). Lucid dreaming as metacognition: Implications for cognitive science. Consciousness and Cognition 3, 246-264.

Kahan, T.L., LaBerge, S., Levitan, L., & Zimbardo, P. (1997). Similari-ties and differences between dreaming and waking cognition: An ex-ploratory study. Consciousness and Cognition 6, 132–147.

Kahn, D., & Hobson, J.A. (1993). Self-organization theory of dreaming. Dreaming 3, 151-178.

Kahn, D., Pace-Schott, E.F., & Hobson, J.A. (1997). Consciousness in waking and dreaming: The roles of neuronal oscillation and neuromodu-lation in determining similarities and differences. Neuroscience 78, 13-38.

Karni, A., Tanne, D., Rubenstein, B.S., Askenasy, J.J., & Sagi, D. (1994). Dependence on REM sleep of overnight improvement of a per-ceptual skill. Science 265, 679-682.

Katz, B., & Miledi, R. (1972). The statistical nature of the acetylcholine potential and its molecular components. Journal of Physiology 224, 665-669.

Kavanau, J.L. (1996). Memory, sleep, and dynamic stabilization of neu-ral circuitry: Evolutionary perspectives. Neuroscience and Biobehavioral Reviews 20, 289-311.

Kerr, N. (1993). Mental imagery, dreams and perception. In Cavallero C., & Foulkes D. (Eds.) Dreaming as cognition, pp 18-37. New York: Harvester Wheatsheaf.

Kitt, C.A., Mitchell, S.J., DeLong, M.R., Wainer, B.H., & Price, D.L. (1987) Fiber pathways of basal forebrain cholinergic neurons in mon-keys. Brain Research 17, 192-206.

Kluger, H.Y. (1975). Archetypical dreams and “everyday” dreams. Is-rael Annals of Psychiatry 13, 6-47.

Kosslyn, S.M. (1994). Imagery and brain: The resolution of the imagery debate. Cambridge, Massachusetts: Bradford/ MIT Press.

Koukkou, M., & Lehmann, D. (1983). Dreaming: The functional state-shift hypothesis. A neuropsychophysiological model. British Journal of Psychiatry 142, 221-231.

Kuiken, D., Neilson, T., Thomas, S., & McTaggart, T. (1983). Com-parison of the story structure of myths, extraordinary dreams, and mun-dane dreams. Sleep Research 12, 196.

Kuper, A. (1979). A structural approach to dreams. Mind 14, 645-662.

LaBerge, S. (1980). Lucid dreaming as a learnable skill: A case study. Perceptual and Motor Skills 51, 1039-1042.

LaBerge, S. (1985). Lucid dreaming. New York: Ballantine Books.

LaBerge, S., & Rheingold, H. (1990). Exploring the world of lucid dreaming. New York: Ballantine Books.

LaBerge, S., & DeGracia, D.J. (1999). Varieties of lucid dreaming ex-perience. In R.G. Kunzendorf, & B. Wallace (Eds.), Individual differ-ences in conscious experience. Philadelphia: John Benjamins Publishing Company. In press.

LaBerge, S., DeGracia, D.J., & Zimbardo, P. (1999). Prolonging lucid dreams. In preparation.

LaBerge, S., Levitan, L., DeGracia, D., & Zarcone, P. (1999). “Out-of-body-experiences,” dreams, and REM sleep. Sleep and Hypnosis. In submission.

Lequerica, A. (1996). Lucid dreaming and the mind-body relationship: A model for the cognitive and physiological variations in rapid eye move-ment sleep. Perceptual and Motor Skills 83, 331-336.

Levitan, H. (1976). The significance of certain catastrophic dreams. Psychotherapy and Psychosomatics 27, 1-7.

Levitan, L. (1994). A fool’s guide to lucid dreaming. Nightlight 6, 1-5. Online: http://www.lucidity.com/NL62.FoolsGuide.html.

Levitan, L., & LaBerge, S. (1991). Other worlds: Out-of-body experi-ences and lucid dreams. Nightlight 3, 1-5. Online: http://www.lucidity.com/NL32. OBEandLD.html.

Liebovitch, L.S., & Tóth, T.I. (1991). A model of ion channel kinetics using deterministic chaos rather than stochastic processes. The Journal of Theoretical Biology 148, 243-267.

Lischka, A. (1979). Erlebnisse jenseits der schwelle: paranormale erfahrungen im wachstand und in luziden traum, bei astralprojektionen und auf seelenreisen. Schwarenburg, Schweiz: Ansata-Verlag.

Llinás, R., & Paré, D. (1991). Of dreaming and wakefulness. Neuroscience 44, 521-535.

Llinás, R., & Paré, D. (1996). The brain as a closed system modulated by the senses. In R. Llinás & P. Churchland (Eds.) The Mind-Brain contin-uum. MA: MIT Press, pp. 1-18.

Llinás, R., & Ribary, U. (1993). Coherent 40-Hz oscillation characterizes dream state in humans. Proceedings of the National Academy of Sciences of the United States of America 90, 2078-2081.

Loewy, A.D. (1990). Central autonomic pathways. In A. Loewy, & K. Spyer (Eds.), Central Regulation of Autonomic Functions, pp. 88-103. New York: Oxford University Press.

Magallón, L. (1991). Awake in the dark: imageless lucid dreaming. Lu-cidity, 10(1-2), 400-405.

Mahowald, M.W., & Schenck, C.H. (1992). Dissociated states of wake-fulness and sleep. Neurology 42, 44-52.

Malamud, A. (1988). Learning to become fully lucid: A program for inner growth. In J. Gackenbach., & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspectives on lucid dreaming, pp. 309-319. New York: Plenum Press.

Mancia, M. (1995). One possible function of sleep: To produce dreams. Behavioral Brain Research 69, 203-206.

Maquet, P., Peters, J.M., Aerts, J., Delfiore, G., Degueldre, C., Luxen, A., & Franck, G.. (1996). Functional neuroanatomy of human rapid-eye-movement sleep and dreaming. Nature 383, 163-166.

Marcot, B.G. (1987). A journal of attempts to induce and work with lucid dreams: Can you kill yourself while lucid? Lucidity Letter 6.

Global Workspace and Dreaming 49

Marr, D. (1983). Vision. San Fransico: W.H.Freeman.

Mavromatis, A. (1987). Hypnagogia. London: Routledge & Kegan Paul.

McCormick, D.A. (1992). Neurotransmitter actions in the thalamus and cerebral cortex and their role in neuromodulation of thalamocortical activity. Progress in Neurobiology 39, 337-388.

McCreery, C. (1973). Psychical phenomena and the physical world. London: Hamish Hamilton.

Meier, B. (1993). Speech and thinking in dreams. In C. Cavallero, & D. Foulkes (Eds.), Dreaming as cognition, pp. 58-76. New York: Harvester Wheatsheaf.

Mellet, E., Petit, L., Mazoyer, B., Denis, M., & Tzourio, N. (1998). Re-opening the mental imagery debate: Lessons from functional anatomy. Neuroimage 8, 129-139.

Mesulam, M.M. (1998). From sensation to cognition. Brain, 121, 1013-1052

Mischel, W., & Shoda, Y. (1998). Reconciling processing dynamics and personality dispositions. Annual Review of Psychology 49, 229-258.

Moffitt, A., & Hoffmann, R. (1987). Basic dream processes. In J. Gack-enbach (Ed.), Sleep and dreams: a source book, pp. 145-186. New York: Garland Publishing Inc..

Moffitt, A., Hoffmann, R., Mullington, J., Purcell, S., Pigeau, R., & Wells, R. (1988). Dream psychology: operating in the dark. In J. Gack-enbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspec-tives on lucid dreaming, pp. 181-220. New York: Plenum Press.

Moffitt, A., Kramer, & Hoffmann, R. (1993). Introduction. In A. Moffitt, M. Kramer & R. Hoffmann (Eds), The functions of dreaming. Albany, NY: State University of New York Press. pp. 1-10.

Montangero, J. (1991). How can we define the sequential organization of dreams? Perceptual and Motor Skills 73, 1059-1073.

Moore, C., & Engel, S.A. (1999). Mind and brain see eye to eye. Current Biology 9, R74-R76.

Morrison, A.R. (1983). Paradoxical sleep and alert wakefulness: Varia-tions of a theme. In M. Chase, & E. Weitzman (Eds.), Sleep disorders: Basic and clinical research, pp. 95-122. New York: SP Medical and Scientific Books.

Murray, E.A. & Bussey, T.J. (1999). Perceptual-mnemonic functions of the perirhinal cortex. Trends In Cognitive Sciences 3, 142-151.

Murri, L., Arena, R., Siciliano, G., Mazzotta, R., & Muratorio, A. (1984). Dream recall in patients with focal cerebral lesions. Archives of Neurology 41, 183-185.

Myers, C.E., Gluck, M.A., & Granger, R. (1995). Dissociation of hippo-campal and entorhinal function in associative learning: a computational approach. Psychobiology 23, 116-138.

Newman, J., & Baars, B.J. (1993). A neural attentional model for access to consciousness: A global workspace perspective. Concepts in Neuro-science 4, 255-290.

Niiyama, Y., Sekine, A., Fushimi, M., & Hishikawa, Y. (1997). Marked suppression of cortical auditory evoked response shortly before the onset of REM sleep. Neuroreport 8, 3303-3308.

Nofzinger, E.A., Mintun, M.A., Wiseman, M.B., Kupfer, D.J., & Moore, R.Y. (1997). Forebrain activation in REM sleep: An FDG PET study. Brain Research 770, 192–201.

Obermayer, K., Sejnowski, T., & Blasdel, G.G. (1995). Neural pattern formation via a competitive Hebbian mechanism. Behavioral Brain Re-search 66, 161-167.

O'Keefe, J. (1979). A review of the hippocampal place cells. Progress in Neurobiology, 13, 419-413.

O'Keefe, J. (1990). A computational theory of the hippocampal cognitive map. Progress in Brain Research. 83, 301-312.

Okuma, T. (1992). On the psychophysiology of dreaming: A sensory image--free association hypothesis of the dream process. Japanese Jour-nal of Psychiatry and Neurology 46, 7-22.

Ophiel, (1982). The art and practice of astral projection. York Beach, Maine: Samuel Wiser, Inc.

Palambo, S. (1978). Dreaming and memory. New York; Basic Books.

Penfield, W., & Perot, P. (1963). The brain record of auditory and visual experience: a final summary and discussion. Brain 86, 595-696.

Petrides, M. (1985). Deficits on conditional associative-learning tasks after frontal- and temporal-lobe lesions in man. Neuropsychologia 23, 601-614.

Pigarev, I.N., Nothdurft, H.C., & Kastner, S. (1997). Evidence for asyn-chronous development of sleep in cortical areas. Neuroreport 8, 2557-2560.

Porte, H.S., & Hobson, J.A. (1996). Physical motion in dreams: One measure of three theories. Journal of Abnormal Psychology 105, 329-335.

Posner, M. (1982). Cumulative development of attentional theory. American Psychologist 37, 168-179.

Posner, M., & Raichle, M. (1994). Images of mind. New York: Scientific American Library.

Posner, M.I., & Tudela, P. (1997). Imaging resources. Biological Psy-chology 21, 95-107.

Posner, M.I., DiGirolamo, G.J., & Fernandez-Duque, D. (1997). Brain mechanisms of cognitive skills. Consciousness and Cognition 6, 267-290.

Price, R.F., & Cohen, D.B. (1988). Lucid dream induction an empirical evaluation. In J. Gackenbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspectives on lucid dreaming, pp 105-134. New York: Plenum Press.

Purcell, S., Moffitt, A., & Hoffmann, R. (1993). Waking, dreaming and self-regulation. In A. Moffitt, M. Kramer & R. Hoffmann (Eds.), The functions of dreaming, pp. 197-260. Albany, NY: State University of New York Press.

Purcell, S., Mullington, J., Moffitt, A., Hoffmann, R., & Pigeau, R. (1986). Dream self-reflectiveness as a learned cognitive skill. Sleep 9, 423-437.

Quartermain, D. (1983). The role of catecholamines in memory process-ing. In J.A. Deutsch (Ed.), The physiological basis of memory. New York: Academic Press.

50 DeGracia

Ravagnati, L., Halgren, E., Babb, T.L., & Crandall, P.H. (1979). Activity of human hippocampal formation and amygdala neurons during sleep. Sleep 2, 161-73.

Rechtschaffen, A. (1978). The single-mindedness and isolation of dreams. Sleep 1, 97-109.

Rechtschaffen, A., Hauri, P., & Zeitlin, M. (1966). Auditory awakening thresholds in REM and NREM sleep stages. Perceptual and Motor Skills 22, 927-942.

Rechtschaffen, A., & Kales, A. (1968). A Manual Of Standardized Ter-minology Techniques And Scoring System For Sleep Stages Of Human Subjects. Brain Information Service/Brain Research Institute, U.C.L.A.

Ridley, R.M., Harder, J.A., & Baker, H.F. (1996). Neurochemical modu-lation of the hippocampus in learning, remembering and forgetting in primates. Neurodegeneration, 5, 467-471.

Rifat, C. (1980). De la structure du rêve à un modèle informationnel de la schizophrénie. Agressologie. Paris:Masson.

Rifat, C. (1997). Sur les personnages de rêve. Rêver 1, 154-156.

Rock, I. (1997) Indirect Perception. Cambridge, MA: MIT Press.

Rogo, D.S. (1986). Leaving the body. New York: Prentice Hall.

Rosa, R.R., & Bonnet, M.H. (1985). Sleep stages, auditory arousal threshold, and body temperature as predictors of behavior upon awaken-ing. International Journal of Neuroscience 27, 73-83.

Rotenberg, V.S. (1992a). Sleep and memory. I: The influence of differ-ent sleep stages on memory. Neuroscience & Biobehavioral Reviews 16, 497-502.

Rotenberg, V.S. (1992b). Sleep and memory. II: Investigations on hu-mans. Neuroscience and Biobehavioral Reviews 16, 503-505.

Rotenberg, V.S. (1996). REM sleep and dreams as mechanisms of the recovery of search activity. In A. Moffitt, M. Kramer & R. Hoffmann (Eds), The functions of dreaming, pp. 261-292. Albany, NY: State Uni-versity of New York Press.

Rotenberg, V.S., & Arshavsky, V.V. (1979). REM sleep, stress and search activity. A short critical review and a new conception. Waking & Sleeping 3, 235-244.

Roth, T., Roehrs, T., Zwyghuizen-Doorenbos, A., Stepanski, E., & Wit-tig, R. (1988). Sleep and memory. Psychopharmacology Series 6, 140-145.

Rupniak, N.M.D., & Gaffan, J. (1987). Monkey hippocampus and learn-ing about spatially directed movements. Journal of Neuroscience, 7, 2331-2337.

Rye, D.B., Saper, C.B., Lee, H.J., & Wainer, B.H. (1987). Pedunculo-pontine tegmental nucleus of the rat: cytoarchitecture, cytochemistry, and some extrapyramidal connections of the mesopontine tegmentum. Jour-nal of Comparative Neurology 259, 483-528.

Salzarulo, P., & Cipolli, C. (1974). Spontaneously recalled verbal mate-rial and its linguistic organization in relation to different stages of sleep. Biological Psychology 2, 47-57.

Schacter D.L. (1976). The hypnagogic state: a critical review of the lit-erature. Psychological Bulletin 83, 452-81.

Schacter, D.L., & Buckner, R.L. (1998). On the relations among prim-ing, conscious recollection, and intentional retrieval: Evidence from

neuroimaging research. Neurobiology of Learning and Memory 70, 284-303.

Schacter, D.L., Norman, K.A., & Koutstaal, W. (1998). The cognitive neuroscience of constructive memory. Annual Review of Psychology 49, 289-318.

Schacter, D.L., & Tulving, E., Eds. (1994). Memory systems 1994. Cam-bridge, Massachusetts: MIT Press.

Schanfald, D., Pearlman, C., & Greenberg, R. (1985). The capacity of stroke patients to report dreams. Cortex 21, 273-247.

Schiller, P.H. (1997). Past and present ideas about how the visual scene is analyzed by the brain. Cerebral Cortex 12, 59-90.

Schneck, J.M. (1968). An evaluation of hypnagogic hallucinations. Psy-chiatry Quarterly 42, 232-234.

Schwartz, W., & Godwyn, M. (1988). Action and representation in ordinary and lucid dreams. In J. Gackenbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspectives on lucid dreaming, pp. 419-427. New York: Plenum Press.

Sherman, S.M., & Koch, C. (1986). The control of retinogeniculate transmission in the mammalian lateral geniculate nucleus. Experimental Brain Research 63, 1-20.

Silberer, H. (1917). Hidden symbolism of alchemy and the occult arts. Jelliffe. S.E (Trans). New York: Dover.

Smith, C. (1996). Sleep states, memory processes and synaptic plasticity. Behavioural Brain Research 78, 49-56.

Smith, C., & Lapp, L. (1986). Prolonged increases in both PS and num-ber of REMS following a shuttle avoidance task. Physiology & Behavior 36, 1053-1057.

Smith, K.N., & Nicolas, A. (1998). Recombination at work for meiosis. Current Opinion in Genetics and Development, 8, 200-211.

Smith, R.C. (1986). Evaluating dream function: Emphasizing the study of patients with organic disease. Journal of Mind and Behavior 7, 397-410.

Smythies, J. (1997). The functional neuroanatomy of awareness: With a focus on the role of various anatomical systems in the control of inter-modal attention. Consciousness and Cognition 6, 455–481.

Snyder, T., & Gackenbach, J. (1988). Individual differences associated with lucid dreaming. In J. Gackenbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspectives on lucid dreaming, pp. 221-259. New York: Plenum Press.

Solms, M. (1995). New findings on the neurological organization of dreaming: Implications for psychoanalysis. Psychoanalytic Quarterly 64, 43-67.

Solms, M. (1997). The neuropsychology of dreams: A clinico-anatomical study. Mahweh, New Jersey: Lawrence Erlbaum Associates.

Spadafora, A., & Hunt, H.T. (1990). The multiplicity of dreams: Cogni-tive-affective correlates of lucid, archetypal, and nightmare dreaming. Perceptual & Motor Skills 71, 627-644.

Squire, L.R. (1994). Declarative and nondeclarative memory: Multiple brain systems supporting learning and memory. In D.L. Schacter, & E. Tulving (Eds.), Memory systems 1994. Cambridge, Massachusetts: MIT Press.

Global Workspace and Dreaming 51

Steriade, M. (1992). Basic mechanisms of sleep generation. Neurology 42, 9-18.

Steriade, M., & Llinás, R.R. (1988). The functional states of the thala-mus and the associated neuronal interplay. Physiological Reviews 68, 649-742.

Steriade, M., Iosif, G., & Apostol, V. (1969). Responsiveness of tha-lamic and cortical motor relays during arousal and various stages of sleep. Journal of Neurophysiology 32, 251-265.

Sterling, P. (1998). Retina. In G.M. Shepard (Ed.), The synaptic organi-zation of the brain, 4th edition, pp. 205-254. New York: Oxford Univer-sity Press.

Stickgold, R. (1999). Sleep: off-line memory reprocessing. Trends in Cognitive Sciences 2, 484- 492.

Strauch, I., & Meier, B. (1992). Den traumen auf der spur. Ergebnisse der experimentellen traumforschung. Ben: Huber.

Suzuki, W.A., & Amaral, D.G. (1994a). Perirhinal and parahippocampal cortices of the macaque monkey: cortical afferents. Journal of Compara-tive Neurology 350, 497-533.

Suzuki, W.A., & Amaral, D.G. (1994b) Topographic organization of the reciprocal connections between the monkey entorhinal cortex and the perirhinal and parahippocampal cortices. Journal of Neuroscience 14, 1856-1877.

Suzuki, W.A. (1996). Neuroanatomy of the monkey entorhinal, perirhi-nal and parahippocampal cortices: Organization of cortical inputs and interconnections with amygdala and striatum. Seminars in the Neurosci-ences 8, 3–12.

Suzuki, W.A., Zola-Morgan, S., Squire, L.R., & Amaral, D. G. (1993) Lesions of the perirhinal and parahippocampal cortices in the monkey produce long-lasting memory impairment in the visual and tactual mo-dalities. The Journal of Neurosciences 13, 2430-2451.

Talland, G. (1965). Deranged memory: a psychonomic study of amnesic syndrome. New York: Academic.

Tart, C.T. (1988). From spontaneous event to lucidity, a review of at-tempts to consciously control nocturnal dreaming. In J. Gackenbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspectives on lucid dreaming, pp. 67-103. New York: Plenum Press.

Tedlock, B. (1987). Dreaming: Anthropological and psychological in-terpretations. Cambridge: Cambridge University Press.

Thaler, S. L. (1996a). Neural networks that create and discover. PCAI, May/June 1996, 16-21.

Thaler, S. L. (1996b). Is neuronal chaos the source of stream of con-sciousness? Proceedings of the World Congress on Neural Networks, WCNN ‘96, Lawrence Erlbaum & Associates.

Theeuwes, J. (1994). Endogenous and exogenous control of visual selec-tion. Perception 23, 429-440.

Tholey, P. (1988). A model for lucidity training as a means of self-healing and psychological growth. In J. Gackenbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspectives on lucid dreaming, pp. 263-285. New York: Plenum Press.

Tholey, P. (1991). Applications of lucid dreaming in sports. Lucidity, 10(1-2), 431-439.

Torda, C. (1969). Dreams of subjects with loss of memory for recent events. Psychophysiology 6, 358-365.

Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & S. Donaldson (Eds.), Organization of Memory. New York, NY: Academic Press. pp. 382-403.

Tulving, E. (1987). Multiple memory systems and consciousness. Hu-man Neurobiology 6, 67-80.

Ungerleider, L.G., & Haxby, J.V.(1994) 'What' and 'where' in the human brain. Current Opinions in Neurobiology 4, 157-165.

Velasco, F., Velasco, M., Cepeda, C., & Munoz, H. (1980). Wakeful-ness-sleep modulation of cortical and subcortical somatic evoked poten-tials in man. Electroencephalography and Clinical Neurophysiology 48, 64-72.

Vogel, G.W. (1975). A review of REM sleep deprivation. Archives of General Psychiatry 32, 749-761.

Von der Malsburg, C. (1996). The binding problem of neural networks. In R. Llinás & P. Churchland (Eds.) The Mind-Brain continuum. MA: MIT Press, pp. 131-146.

Weinberger, D.R. (1993). A connectionist approach to the prefrontal cortex. Journal of Neuropsychiatry 5, 241-253.

Willingham, D.B. (1998). A neuropsychological theory of motor skill learning. Psychology Reviews 105, 558-584.

Wilson, J.R., Friedlander, M.J., & Sherman, S.M. (1984). Fine structural morphology of identified X- and Y-cells in the cat's lateral geniculate nucleus. Proceedings of the Royal Society of London. Series B: Biologi-cal Sciences 221, 411-436.

Winograd, T. (1972). A program for understanding natural language. Cognitive Psychology 3, 1-191.

Witter, M.P., Groenewegen, H.J., Lopes da Silva, F.H., & Lohman, A.H. (1989). Functional organization of the extrinsic and intrinsic circuitry of the parahippocampal region. Progress in Neurobiology, 33, 161-253.

Worsley, A. (1988). Personal experiences in lucid dreaming. In J. Gack-enbach, & S. LaBerge (Eds.), Conscious mind, sleeping brain: Perspec-tives on lucid dreaming, pp. 321-341. New York.: Plenum Press.

Worsley, A. (1991). Dream lucidity induction and control. Lucidity, 10(1-2), 33-36.

Yates, J. (1985). The content of awareness is a model of the world. Psy-chological Review 92, 249-284.

Zeki, S. (1997). The color and motion systems as guides to conscious visual perception. Cerebral Cortex 12, 777-809.